Cognitive tasks materials and procedure

Visual Declarative Learning Ability (Visual-Spatial Associations). The materials for the administration of all declarative memory tasks and the measure of the phonological loop were part of the PROMEA battery (Vicari, 2007). For visual

declarative memory they included a set of A4 full-size color pictures of 16 familiar objects, 16 pictures of the same objects occupying a random position in an A4 four- spaced grid, one A4 grid with no pictures, and a booklet to record the child's scores. The task probed the retention of visual-spatial associations immediately and after a delay, and was administered individually according to the battery's manual. It started by showing the children a practice stimulus and explaining that they would have to memorize the picture position in the grid and would be asked to show the position using the empty grid immediately afterwards. After the purpose of the task was clarified with the practice stimulus, a full sequence of 15 A4 grids with pictures was shown, 5 seconds per stimulus.

Subsequently, the children were given the empty grid they would use to indicate the positions and shown the full set of 15 full-size pictures one by one. The children were assigned one point for each picture position in the grid they were able to recollect correctly. No feedback was given on the correctness of single picture-

position pairings but the researcher told the children how many matches they got wrong at the end of each series. The whole process was repeated three times.

Unbeknownst to the participants, a final delayed recall test was performed 15 minutes after the third immediate test.

Verbal Declarative Learning Ability (Short Story). The materials included a short story of 58 words divided into 28 information units and a form where the version of the story provided by the child in the recall trials could be noted down. The

children were asked to listen to a short story and told they would have to repeat it as precisely as they could immediately afterwards. As the children recounted the story the researcher ticked the information units that were remembered and noted down any different words used or changes in the order of the events. Unbeknownst to the

participants, they were asked to recall the story a second time after 15 minutes and their performance was assessed in the same way.

Declarative Learning Ability: Task scoring. For declarative learning ability three scores were obtained: (a) visual-spatial declarative memory (from the visual- spatial association task), (b) verbal declarative memory (from the short story task), and (c) a global score of declarative memory. All scores were obtained applying the formulas provided in Vicari (2007) and using the norms provided, i.e. selecting the values corresponding to the raw scores adjusted for the child's age (from 7 years 0 months to 7 years 11 months; from 8 years 0 months to 8 years 11 months; and from 9 years 0 months to 9 years 11 months). The use of the formulas for calculating visual and verbal declarative memory was necessary for standardization in accordance with the manual of the tests.

For visual-spatial declarative memory, given that TPo was the sum of positions recalled in the three trials and Rec1 was the number of position recalled in the delayed trial, the formula applied was (Vicari, 2007, p. 37):

s1 = (adjusted TPo + adjusted Rec1) / 2 visdecl score = (s1 * 5) + 50

For verbal declarative memory the score assigned one point for each

information unit accurately recalled in the immediate recall of the story, repeating the procedure for the delayed recall. Accurate recall included when synonymous words were used but excluded recalling accurate information units in the wrong logical

order. Given that Imm was the number of information units recalled in the first trial and Rec2 the number of information units recalled in the delayed trial, the formula was:

s2 = adjusted [(Imm + Rec2) / 2] verbdecl score = (s2 * 5) + 50

Finally, the global measure of declarative memory based on the available measures was calculated as follows:

s3 = (adjusted TPo + adjusted Rec1 + s2) / 3 totdecl = (s3 * 5) + 50

Phonological Loop (Nonword Repetition). The materials included a list of 40 nonwords of different length in a random order: 10 words with two syllables, 10 words with three syllables, 10 words with four syllables, and 10 words with five syllables. Half of each set of words has high phonological similarities and half low phonological similarity with Italian words. For all words the stress falls on the penultimate syllable (the most frequent stress pattern in Italian). The task was

administered by the researcher according to the manual, reading the words to the child one at a time whilst holding a light sheet of paper in front of her mouth so that lip movements would not provide visual cues for pronunciation.

Phonological Loop scores. Standardized z-scores were also calculated for the nonword repetition task providing a measure of the phonological loop. In the nonword repetition task a raw score was obtained assigning a point for every word repeated correctly, and zero points if any pronunciation errors were made on an item

(inaccurate phonemic recall). and the final measure of phonological memory was the standardized raw score thus obtained. According to the administration manual, correct pronunciation was subjectively assessed by the researcher, an Italian native speaker

with no diagnosed hearing impairment, who tested all children. The assessment of the accuracy of the non-word repetition task was done immediately.

Procedural Learning Ability (Alternate Serial Reaction Time Task). In a serial reaction time task participants are usually asked to immediately react to the changes in position of an on-screen target by pressing the corresponding buttons on a controller/input box. The alternating serial reaction time task (ASRT) was obtained modifying a version of an E-Prime serial reaction time task (SRT) originally developed in Lum (2010). The main difference between the ARST and the SRT involves the type of sequence the participants are expected to learn as a result of the exposure to the stimuli. In the SRT each item in a sequence of stimuli in nonrandom blocks belongs to the pattern sequence, whilst in the ASRT the pattern sequence is 'concealed' by an alternation of random and pattern positions. The ASRT presents important advantages compared with the SRT. First of all, it provides a continuous measure of learning, whereas learning in the SRT is assessed only once at the end of the task comparing reaction times on a random block to reaction times on a sequence block. Secondly, as the training sequence in the ASRT includes both pattern and random stimuli, it provides the opportunity to differentiate between sequence learning and general motor skill learning without the need to take additional motor skill

measures. Finally, as a number of studies have consistently shown, sequences in ASRTs are unlikely to be learnt explicitly, as the pattern sequence is usually not reported and/or identified even after extended practice (Hedenius, 2013; Howard et al., 2004).

The task was administered with an ASUS X553M laptop computer and headphones. The children inputted their responses using a game controller that could be configured for both right-handed and left-handed use (iBUFFALOTM Classic USB

Gamepad BSGP801). The use of the game controller and the task interface were aimed at creating an involving and child-friendly gaming environment (Lum, 2010). The task consisted in pressing one of four buttons in the game controller

corresponding to the position a smiley would appear in on screen (Figure 6.5).

Figure 6.5. Alternate Serial Reaction Time Task.

To further discourage the use of explicit learning strategies (and unlike previous version of the ASRT task, e.g., Hedenius, 2013), the task deployed here did not provide indirect feedback on incorrect trials by blocking the transition to the next stimulus until the correct button was pressed. Rather, independently of whether the response to a given stimulus was correct, the next stimulus was immediately presented.

The ASRT task began with on-screen instructions and a series of training trials designed to familiarize the player with the task and the controller. After that there were 8 experimental blocks (80 trials each), with the possibility for the participant of taking brief self-managed breaks between blocks. The blocks and the trials in each block were administered to all participants in the same order, and each block was preceded by 5 warm-up trials. The last screen in each block gave the participants feedback about their accuracy (percentage correct) and about the speed of their

performance in the block. In the experimental trials the sequence alternated a fixed position with a random position according to the pattern 1 r 2 r 4 r 3 r (Hedenius, 2013).

The children were told to play the game trying to press the correct controller button as fast as they could. They were also told to check the scores at the end of the block and try to keep an accuracy score around 92% (Howard et al., 2004), whilst at the same time trying to improve their speed score as the game progressed.

Procedural Learning Ability scores. Procedural learning ability was

operationalized as learning of the fixed pattern in the alternating stimuli sequence the participants were exposed to. The ASRT task allowed to obtain three measures of procedural learning ability. The first measure was based on the reaction times, the second on (in)accuracy, and the third was a composite of the previous two. For the RT measure, pattern learning was operationalized as the finding of increasingly shorter RTs for pattern trials compared to random trials as training progressed. For the

(in)accuracy measure, pattern learning was operationalized as an increasing number of errors on random trials compared to pattern trials as training progressed. The ASRT data were first reduced in Excel excluding practice trials, warm-up trials, incorrect trials, and trials that were the final element in 'trills' (e.g., 212) or in 'repetitions' (e.g., 222), as recommended in Hedenius (2013) and Howard et al. (2004).

For the RT-based measure, the median RT values in milliseconds were calculated separately for sequence trials and nonsequence trials for each participant's block (Bennet at al., 2011, Hedenius, 2013). In both data subsets the scores from Block 1 to Block 4 and the scores from Block5 to Block 8 were averaged obtaining an A and a B score respectively. The difference between A and B (RT Gain) reflected the change in reaction times from the first half to the second half of the training. To obtain

a final measure of procedural learning ability based on RTs, RT Gains from

nonsequence trials were subtracted from RT Gains from sequence trials, with higher positive differences indicating better sequence learning.

In the case of the (in)accuracy measure the same reduced dataset was used, but this time it also included incorrect items. To obtain a measure of procedural learning ability based on (in)accuracy, the number of inaccurate responses (errors) was

calculated for each participant's block and then averaged across blocks (Bennet at al., 2011, Hedenius, 2013). For each participant the difference between the average number of errors in nonsequence trials and the average number of errors in sequence trials provided a measure of sequence learning, with larger positive differences indicating better sequence learning. Finally, a composite measure of procedural learning ability (Proc) was obtained standardizing and then averaging the two components.