## Introduction

Learning to read involves establishing connections between graphemes (printed words) and phonemes (sounds). Generally, children are taught to read starting from 6 years of age (Italian mainstream) and the full mastery of reading requires some years of practice and instruction. Once completely automatized, reading becomes fluent and does no longer require conscious control. However, the 3–7% of the school population struggles in automatizing the reading process. Indeed, these children suffer from Developmental Dyslexia (henceforth DD). In Italy, the percentage ranges from 1.5 to 5%1,2,3,4. Children with DD turn out to read slower and less accurately than children of equal age who received the same amount of education. The Diagnostic and Statistical Manual of Mental Disorders (DSM-5) describes Developmental Dyslexia as “a pattern of learning difficulties characterized by problems with accurate or fluent word recognition, poor decoding, and poor spelling abilities”5 (p. 67) in children with average intellectual abilities, without mental or neurological disorders, proficient in the language of academic instruction and provided with adequate educational instruction. Developmental Dyslexia is thus included among the specific learning disorders.

Among the numerous theories seeking to explain Developmental Dyslexia, the Phonological Theory, according to which the core problem is due to phonology related processes6,7,8, enjoys wide agreement. Children, who turn out to be affected with DD usually show a phonological deficit before they learn to read and its severity is usually predictive of the variation in the severity of the reading deficit. By consequence, children with DD struggle with reading accurately and fluidly and experience particular difficulties in phonological awareness tasks (explicit phonology), in phonological processing tasks and in tasks demanding rapid automatized naming (implicit phonology). As a quick overview, children with DD have been found to perform more poorly than younger reading age matched children in the selection of the word out of a sequence of four words that did not rhyme9; in syllable segmentation10; in nonwords repetition11, in rapid automatized naming (RAN)12.

Adults were tested in three conditions: the Unstressed condition (condition 1), which was a plain metronome pulse; the Stressed condition (condition 2), which consisted of an alternation of strong and weak beats; the Unpredictable condition (condition 3), which served as a control condition since it consisted of an unpredictable sequence of beats. Children were only tested in the Unstressed condition (condition 1) and the Unpredictable condition (condition 3). Details for each conditions are reported in Fig. 4. For each condition, ten WB-IB couples were presented, which we will refer to as Repetitions.

Regarding the measurements, the accepted behaviour for simple mean reaction time is a delayed response of about 150 ms for sound stimuli70,71; any positive error smaller than 150 ms could be interpreted as a measure of some anticipatory behavior. When asking to tap in synchrony with a simple sequence of auditory tones, there is a systematic tendency for tap responses to automatically anticipate the signal by minus 30–50 ms72, usually referred to as negative mean asynchrony (NMA). The task we adopted (warning-imperative) is also clearly distinct from a reaction time task. On the one hand, we expected typically developed individuals to tap at the same time with the IB or slightly before as a consequence of the NMA64,73. On the other hand, if individuals with DD suffer from an impairment in anticipatory abilities, they may have trouble to prepare themselves and decide when to tap at the IB. We planned to measure the timing error (calculated as the difference between the observed time, namely the time of the key press on the IB and the expected time, namely the IB time) and the individual consistency (the standard deviation of timing error computed on each set of ten responses for each participant). When referring to our experimental conditions, we expected to find group differences when measuring the timing error between individuals with and without DD in the Unstressed and Stressed Conditions for adults; only in the Unstressed condition for children (since only this condition was tested), but we did not have any specific prediction beforehand on whether DD participants would over-anticipate or would delay the estimation of the occurrence of the IB. We also predicted DD participants to be individually less consistent than the control participants. Importantly, if the group differences were due to the fact that individuals with DD are generally slower than typically developing individuals, we would expect group differences also in the Unpredictable condition, i.e., individuals with DD should respond on average more slowly than controls. On the contrary, if individuals with DD have only poor anticipatory skills, we do not expect any group difference in the Unpredictable condition, as beats being unpredictable, there is no basis for building a temporal representation to exploit during the testing phase.

One ingredient of our task, needed also in the tapping task, is perceptual synchronization, a skill that develops over years and reaches adult-like performance at the age of 6–774. In addition, our task involves a double inhibition component: first, participants needed to refrain from tapping to every beat; second, they have to inhibit their tapping to WB and tap to the IB. That is, our task also involved the inhibition component of executive function skills, which are known to develop over the year even up to 12 years of age75. Therefore, we expected children (with and without DD) to display a higher NMA.

We also tested reading abilities (speed and accuracy) both in adults and children, to test potential correlations between anticipation skills and reading abilities. In addition, we measured basic motor skills of DD adults through the Pegboard battery. The fine motor abilities of children were not tested since previous findings in the literature showed that manual dexterity of children with DD does not differ from that of their peers41.

## Results

Generalized Linear Model (GLM) analyses with Group (DD, TD) and Gender (M, F) as between-subject factor on reading time (syllable/second) and error scores revealed a main effect of Group for the error scores, F(1, 35) = 39.51, p < 0.001, η2p = 0.53, and for reading time, F(1, 35) = 94.4, p < 0.001, η2p = 0.73. Adults with DD were slower and made more errors than TD adults.

GLM analyses were conducted on error scores and reading time (syllable/second) with Group (DD, TD) and Gender (M, F) as between-subject factors, and Task (Words, Non-Words) as within-subject factor. A significant main effect of Group, F(1, 43) = 60.4, p < 0.001, η2p = 0.58 was found for the error score, as DD children made more errors than TD children. For the reading time, the interaction Task by Group was significant, F(1, 43) = 37.6, p < 0.001, η2p = 0.46, as well as the main effects of Group, F(1, 43) = 77.6, p < 0.001, η2p = 0.64, and of Task, F(1, 43) = 153.2, p < 0.001, η2p = 0.78. As expected, DD children read slower than TD children; the interaction was due to the fact that TD children read words faster than non-words (whereas DD children were slow both when reading words and when reading non-words).

### Performance on the Purdue Pegboard Battery (adult participants)

Two separate GLM analyses were performed on the scores of the first three sub-tasks (Right Hand (RH), Left Hand (LH), Both, treated as within-subject factor) and on the last sub-task (Assembly), with Group (DD, TD) and Gender (F, M) as between-subject factor. The first analysis showed a main effects of Group, F(1, 35) = 4.93, p < 0.05, η2p = 0.12, as DD participants inserted fewer pins than TD participants, and a main effect of sub-task F(2, 70) = 4.93, p < 0.001, η2p = 0.46, as the skill decreased from RH to LH. Neither main effects nor interactions emerged from the second analysis on Assembly score. Mean and standard error of the score of each sub-task are reported in Table 1.

#### Pre-processing of data

In order to detect outliers in each set of repeated responses for the Unstressed (1) and Stressed (2) Conditions, we filtered the data of each participant using the Median Absolute Deviation (MAD)76 computed on the ten repetitions. Then we calculated the z point corresponding to each response xi:

$${z}_{i}=\left|\frac{0.6745 \cdot ({x}_{i}-M)}{MAD}\right|$$

where M stands for median. When z was > 2, the data point was replaced by the median. As a result, the 4% of the adult responses and the 4.4% of children responses were substituted. We did that because the warning-imperative task is very demanding and a momentary drop of attention can produce clearly aberrant responses. Thus, this preliminary procedure of filtering data was meant to exclude these very few meaningless responses (for instance, errors greater than 750 ms or smaller than − 250 ms). The data of the Unpredictable condition were left unfiltered given their expected great variability. Afterwards, also the presence of possible outliers among the participants was checked. Participants with a mean error (computed on the Unstressed (1) and Stressed (2) Conditions) outside the interval delimited by ± 3SD of participants’ mean were considered outliers. As a result, one participant of the adult TD group was discarded. Two dependent variables were considered for the analysis: the timing error and the individual consistency.

### Timing error

In order to determine whether the timing of tapping was synchronous with the occurrence of the IB beat, the timing error was calculated as the difference between the observed time, namely the time of the key press on the IB and the expected time, namely the IB time. Thus, positive errors represent delayed responses; negative errors represent anticipated responses; zero errors represent flawless simultaneity. Timing error was calculated for each condition.

Figure 1 displays the performance of typical adult participants. As displayed in Fig. 1, the TD participant showed good predictive skills. In the Unstressed (1) and the Stressed (2) Conditions, adults with TD responded in synchrony with the IB (in sporadic cases the response had few milliseconds of anticipation). On the contrary, adults with DD tended to give a delayed response to the IB and tended to be very variable. In the Unpredictable condition (3), which is meant to be a control condition, participants of both groups were not able to anticipate the IB (as there was no regularity). Both the DD and the TD participants tended to respond after having heard the IB. This finding supports our hypothesis that individuals with DD struggle in temporal predictions and are not merely slower than controls.

GLM analyses on the timing error were carried out for each condition, with Group (DD, TD) and Gender (M, F) as between-subject factors, and Repetition as within-subject factor. Results are shown in Fig. 2 and reported in Table 2.

In the Unstressed condition (Condition 1), a main effect of Group was found, F (1, 36) = 5.89, p < 0.05, η2p = 0.14. As illustrated in Fig. 2, TD participants were synchronous or slightly anticipated the IB whereas participants with DD were delayed with reference to the IB timing. In the Stressed condition (Condition 2), we also found a main effect of Group, F(1, 36) = 6.91, p < 0.05, η2p = 0.16. As displayed in Fig. 2, participants with DD showed a similar tendency as for the Unstressed condition. Repetitions was also significant, F(9, 315) = 4.39, p < 0.001, η2p = 0.11. Bonferroni post-hoc comparisons showed that the timing error of the first repetition was bigger than the timing errors of the third, fourth and fifth repetition. The interaction, Group x Repetition was significant, F(9, 315) = 2.27, p < 0.05, η2p = 0.06, though Bonferroni post-hocs revealed no significant comparisons. The interaction Gender × Repetition was also significant, F(9, 315) = 2.35, p < 0.05, η2p = 0.06, although again no statistical significant post-hoc comparisons were found. In the Unpredictable (3) condition, no significant difference was found.

#### Children

The average performance of the Unstressed (condition 1) and the Unpredictable condition (condition 3) is shown in Fig. 3 and Table 2. The results showed that children’s behavior is similar to the adults’ one. The same GLM analyses on the timing error used for adults were carried out for children. In the Unstressed condition (condition 1), a main effect of Group was found, F(1, 43) = 20.8, p < 0.001, η2p = 0.33. As illustrated in Fig. 3, TD participants anticipated the IB whereas participants with DD were delayed with reference to the IB timing. The interaction Group by Gender by Repetition was significant, F(9, 387) = 2.1, p < 0.05, η2p = 0.05. This two-ways interaction is difficult to interpret given the weak η2p and that Bonferroni post-hocs revealed only 4 significant comparisons out of 780. In the Unpredictable condition (condition 3), no significant difference was found except the interaction Group by Gender, F(1, 43) = 4.4, p < 0.05, η2p = 0.09, though Bonferroni post-hocs revealed no significant comparisons.

### Individual consistency

This measure characterizes an individual’s coherence of IB tapping response across the ten repetitions. Individual consistency was measured by calculating the standard deviation of timing error computed on each set of ten responses for each participant.

The GLM analysis on the Unstressed condition (condition 1) with Group (DD, TD) and Gender (M, F) as between subject factors revealed a main effect of Group, F(1, 35) = 26.4, p < 0.001, η2p = 0.43, as participants with DD were less consistent within their tapping response than TD participants (71 vs. 35 ms). Gender was also significant, F(1, 35) = 6.1, p < 0.05, η2p = 0.15, as female participants were less consistent than male participants (61 vs. 44 ms). As for the Stressed condition (condition 2), the GLM analysis with Group (DD, TD) and Gender (M, F) as between subject factor revealed a main effect of Group, F(1, 35) = 18.1, p < 0.001, η2p = 0.34. Similarly, to the Unstressed (1) condition, DD participants were less consistent within their tapping response than TD participants (78 vs. 42 ms). We also found a main effect of Gender, F(1, 35) = 13.1, p < 0.001, η2p = 0.27, due to female participants being less consistent than male participants (75 vs. 45 ms). The interaction Gender × Group was also significant, F(1, 35) = 5.50, p < 0.05, η2p = 0.13. Female participants with DD turned out to be significantly less consistent than TD female participants (103 vs. 47 ms), whereas no difference was found between DD and TD male participants (53 vs. 37 ms). Finally, the GLM analysis on the Unpredictable condition (condition 3), with Group (DD, TD) and Gender (M, F) as between subject factor did not reveal any significant difference.

#### Children

The same GLM analyses on consistency (Unstressed (1) and Unpredictable (3) conditions) run on adult data were carried out on child data. A main effect of Group was found both for the Unstressed (1) and the Unpredictable (3) conditions (F(1,43) = 17.7, p < 0.001, η2p = 0.29 and F(1,43) = 20.8, p < 0.001, η2p = 0.33) indicating that DD children were generally less consistent than TD children. No further significant effect or interaction was found.

### Correlation analyses (both adult and child participants)

A correlation matrix among timing error, individual consistency, reading speed, reading errors and Pegboard scores was computed in order to evaluate the relations among different variables. As timing errors we considered the means obtained by collapsing the ten repetition of each participant. Table 3 reports the correlation matrix for adults. Table 4 reports the correlation matrix for children.

To obtain more information from the adult correlation matrix, we run partial correlations, controlling for motor skills (Pegboard Right; Pegboard Left; Pegboard Both; Pegboard Assembly). The results, reported in Table 5, showed that the individual consistency score of Unstressed condition (1) negatively correlated with reading speed and positively correlated with reading errors. No correlation is found between timing error and reading speed when controlling for motor skills.

## Discussion

The current study explored the hypothesis that individuals with Developmental Dyslexia (DD) suffer from an inefficient anticipatory timing mechanism. A growing body of research has observed that individuals with DD have difficulties in time perception in various domains32,43,66,77. At the same time, another strand of literature has uncovered associations between musical skills—especially aspects related to rhythm—and phonological awareness in TD individuals36,38,48,49,50. Individuals with DD have also been reported to benefit from training in sensorimotor synchronization to a pulse delivered by a metronome51, and more generally from music training52. Previous studies focusing on sensorimotor synchronization showed that the sensorimotor coupling is generally well-preserved in the dyslexic population (though sometimes dyslexics are less precise than controls in tapping), but that neural rhythmic entrainment is atypical in individuals with dyslexia41,65,66,67,68.

Children’s results replicated those found with adults. In the Unstressed condition, control children anticipated the IB and tapped generally 30–50 ms in advance, displaying a clear negative mean asynchrony. Children with DD were systematically delayed. As adults with DD, they tried to anticipate the imperative beat, but they had a poor performance. Similarly to adults, children with DD, were less consistent in tapping to the IB than their controls, indicating once again that temporal prediction is challenging for them. In the Unpredictable condition (3), our control condition, no group difference was observed in children as well.

When comparing control children and control adults, it emerged that control children’s negative mean asynchrony was higher than that of adult controls. In the literature, sensorimotor synchronization skills reach adult-like performance at the age of 6–774. Although our children were 9 years old, it is plausible to think that they did not match completely adults’ performance, since our task was not a purely sensorimotor synchronization task, as it involved a double inhibition component: first, participants had to inhibit their tendency to tap to every beat; second, they had to inhibit their tapping to the warning beat and tap to the imperative. We know that inhibition is a complex construct of executive function that develops over the years and some of its components are still developing at age 1275.

In conclusion, DD participants showed a tendency for tapping after the occurrence of the IB both in the Unstressed and in the Stressed conditions (this latter only adults), but their response was lower than the average expected reaction time.

Our results point to the conclusion that well-compensated adults and children with DD have difficulties with the anticipation of temporal events, despite the high predictability of the stimulus, since in our task we could measure the response accuracy to the highly predictable IB after the warning of the WB. Importantly, the no significant group difference in the control condition ruled out the possibility that individuals with DD are generally delayed in their responses. In this condition, the occurrence of the IB was unpredictable, and no regularity could be exploited in order to predict the incidence of the IB.

Our results call into question Tallal’s rapid auditory processing theory22,23, according to which, individuals with DD are not able to integrate sensory information that converges in rapid succession in the central nervous system. The theory was based on findings showing that children with DD could discriminate basic acoustic information (tones of 75 ms) on a par with typically developing children when the inter-stimulus-interval (ISI) was 428 ms, but not when the ISI was 150 ms. However, in our experiment dyslexics struggled to anticipate the forthcoming beat even though the onset-to-onset interval was 750 ms (80 bpm, ISI 550), thus suggesting that it is not frequency to be the source of the difficulties.

As far as we know, this is the first work which explicitly investigates anticipatory skills in Developmental Dyslexia. As such, more than answering questions, this paper leaves many alternatives open. In our task, we tested the ability to anticipate an upcoming input by producing a response to the imperative beat. In turn, this was possible based on a regularity that could be extracted from a perceived regular string. The present experiment leaves unanswered whether the difficulty is in the perception of the rhythm or in the production of rhythm or both, even though previous results suggest that the timing representation itself is deficient in individuals with DD67,68. A deficient timing representation might impact anticipatory abilities—as highlighted by our results—while effecting less synchronization abilities. However, as discussed earlier, the response of individuals with DD was lower than a reaction time. This is compatible with the idea that DD participants attempted to anticipate and they were not merely responding after they heard the IB. However, they were not as good as control individuals. This could suggest that they were able to extract an abstract representation of the regularity—probably assisted by the ease of the task—but were not able to use this representation efficiently.

Individuals with DD experience a wide range of difficulties; nevertheless, they mainly struggle in accurate and/or fluent word recognition and with poor spelling and decoding abilities that are the result of a deficit in the phonological component of language. Future studies should investigate the possible connections between anticipation and phonology, which is the domain of greater difficulties attested in individuals with DD, by relating phonological awareness abilities with anticipatory skills, to better understand how a damage on anticipatory abilities could also impact the phonological domain. All these questions are of great interest and motivate the need for more fine-grained future investigation to better characterize the source of difficulty experienced by individuals with DD.

## Methods

### Participants

Sixteen participants diagnosed with Developmental Dyslexia (DD) (mean age = 22.75; SD = 2.83, 6 female) and 23 control participants (TD) (mean age 24.78, SD = 5.93, 10 female) were tested. The two groups did not differ in age (p = 0.22). All participants were born in Italy, were Italian monolingual speakers, used Italian as their first oral and written language and were students at the University of Milano-Bicocca. They were all right-handed and with normal hearing. DD participants were recruited through the University Learning Disabilities Centre. They were diagnosed for DD following the Italian standard criteria: scores at or below the 5th percentile (or 2 SD from the mean with respect to age) in two out of six measures, these measures being reading speed and accuracy in reading words, pseudowords, and text83; absence of neurological and sensorial disorders; IQ within 1 SD from the mean; adequate socio-cultural opportunities. Therefore, adult participants with DD included in this study received the diagnosis by an authorized clinical institute who followed these criteria prior entering the university. The diagnosis of DD participants was further confirmed at the end of high school (based on reading text). TD participants had no neurological, psychiatric and auditory deficits and no learning disabilities. On the basis of a preliminary interview with the participants aimed at ascertaining their musical competence, one DD participant was excluded. This participant had been played an instrument for 11 years at a semi-professional level. We excluded her, as musical training may enhance rhythmic abilities52 and, if our hypothesis is correct, predictive skills.

#### Children

Eighteen participants diagnosed with Developmental Dyslexia (DD) (mean age = 9.84; SD = 1.0, 9 female) and 29 control participants (TD) (mean age 9.67, SD = 0.73, 13 female) were tested. The two groups did not differ in age (p = 0.70). All participants were born in Italy, were Italian monolingual speakers and used Italian as their first oral and written language, were all right-handed and with normal hearing. Participants with DD were recruited from three Italian clinical institutions: the Developmental Neurology Unit of the Neurological Institute Carlo Besta, the Policlinico of Milan and the Centro di Psicomotricità of Lodi. Dyslexia was diagnosed in accordance with the Italian standard criteria by the qualified teams of each institution involved in the study, which also determined that the DD children had no psychological, neurological or auditory problems, nor did they have Developmental Coordination Disorders. Italian standard criteria for the diagnosis of dyslexia are as above: scores at or below the 5th percentile (or 2 SD from the mean with respect to age) in two out of six measures, these measures being reading speed and accuracy in reading words, pseudowords, and text83; absence of neurological and sensorial disorders; IQ within 1 SD from the mean; adequate socio-cultural opportunities. The TD children were recruited from three different schools in Milan and Verona (north of Italy). The tenets of the Declaration of Helsinki84 were observed and the study was approved by the Ethics Committee of the University of Milano-Bicocca (protocol 0012673/13 for adults, protocol 0010172/13 for children) and also by the one of Istituto Neurologico Carlo Besta (275/2012 for children).

Before each testing session, adult participants were explained about the purpose and the procedures of the study by the experimenter and signed an informed consent. As for child participants, their parents were fully informed about the study and signed informed consent.

### Materials

The Prova di velocità di lettura di brani per la Scuola Media Superiore85 was used to assess reading proficiency. Participants were asked to read out loud the first of the two texts included in the test battery (the text was “Funghi in città”, from Marcovaldo86). Reading speed and error score were considered as dependent variables. Reading speed was scored in syllables per second (571 syllables divided by reading time). Error score corresponded to the number of the mistakes made, counted as follows: 1 point for each word read incorrectly (irrespective of the errors made in the same word); 0.5 point for: shift of accent (e.g., cittadìno → cittàdino); self-correction; lexical substitution (e.g. mutamenti → cambiamenti); same error reiterated in a word presented more time in the text (e.g. di → da, da; di → da) ; uncertainty (e.g., desideri → desi-desideri). The maximum time participants were given to complete the test was 4 min.

#### Testing children

Part 2 and 3 of the Batteria per la valutazione della Dislessia e della Disortografia evolutiva-2, DDE-288 were used to measure single word reading. Children were asked to read aloud four lists of words and three lists of non-words that adhere to the phonotactic constraints of Italian. The dependent variables considered were reading speed and errors’ score. Reading speed was expressed in syllables per second (281 for the words and 127 for non-words divided by reading time). The errors’ score corresponded to the number of words and non-words read incorrectly. Self-correction was not counted as a mistake.

We designed an experiment based on the Warning-Imperative Paradigm69, in order to assess participants’ ability to generate timing prediction according to a given rhythm and to prepare the motor act. During the familiarization phase, participants listened to a regular sequence of beats. The same sequence was heard in the testing phase, with the exception that randomly there were couples of adjacent beats that were different from the others. In these couples, the first beat was the warning (henceforth WB) and the second was the imperative (henceforth IB) beat. The WB was predictive about the timing of the IB, i.e., when the IB was to come. Three different conditions were always presented in the following order (see Fig. 4):

• Unstressed condition (1) It was a plain metronome rhythm, which had a reference tempo of 80 bpm (onset-to-onset intervals of 750 ms). The beats were 440 Hz pure tones with 8 ms rise and fall times and 200 ms steady-state duration. Each couple of WB-IB was randomly presented in a train of 8 beats at random (6 basic tones and a WB-IB couple tones).

• Stressed condition (2) It was similar to the Unstressed condition with the only difference that it consisted of an alternation of strong and weak beats. The strong beats were 440 Hz pure tones with 8 ms rise and fall times and 200 ms steady-state duration. The weak beats were 440 Hz pure tones with 4 ms rise and fall times and 100 ms steady-state duration. The intensity of the weak tone was half of the intensity of the strong one. Beats were presented with onset-to-onset intervals of 750 ms. As for the Unstressed condition, WB-IB couples were randomly inserted in sequences of 8 beats (6 basic tones and a WB-IB couple tones).

• Unpredictable condition (3) It consisted of an unpredictable sequence of beats. Auditory stimuli were identical to those used in the Unstressed condition. Sounds were presented with a mean onset-to-onset intervals of 750 ms plus an error randomly chosen in an interval of 250 ms. As for the previous conditions, WB-IB couples were randomly inserted in sequences of 8 beats. Given the unpredictable timing, this pattern served as control condition.

Both adults and children were administered this task: adults were tested in all three conditions, whereas children were tested only in the Unstressed condition and the Unpredictable condition, for reason of time and for avoiding loss on attention.

For each condition, the dependent measures were the timing error (the difference between the observed time, namely the time of the key press on the imperative beat and the expected time, namely the imperative time), and the individual consistency (measured as the variability of repetitions).

Adult participants were tested individually in a quiet room at the University of Milano-Bicocca, whereas children were tested at the diagnosis centers or at their schools. The total testing session lasted about 30 min, with breaks whenever required. The procedure of the anticipatory timing task is displayed in Fig. 5.

For each condition, participants were first presented a familiarization phase (indicated by a red dot) followed by a warning-imperative phase (indicated by a green dot) (Fig. 5). During the familiarization phase, participants were asked to listen attentively to the given meter (no action was required). During the warning-imperative phase, participants were presented the same rhythmic pattern they were previously familiarized with and were asked to tap to the left mouse button simultaneously with the IB. The interval between the two phases was 3 s. The WB and IB beats were randomly distributed throughout the sequence. In all the conditions, they were obtained by adding the first overtone (880 Hz) to the basic sound used to present the rhythmic pattern. Thus, the WB and IB beats were easily recognizable. In the familiarization phase, one train was presented 2 times per condition; in the warning-imperative phase, one train was presented ten times per condition. In the analyses, the ten WB-IB couples for each condition are referred as Repetitions. Before the test session, participants were trained to tap the left mouse button in response to the IB by using a shorter version of the experiment in which only the Unstressed (1) condition was presented. In this version, one train was presented 2 times during the familiarization phase and 5 times during the warning-imperative phase. The experiment was carried out using MATLAB (R2013a) and PsychToolbox_389. All sounds were played via loudspeakers.