Passive exposure

Here I will refer to passive exposure as the phase of training during which the participants listened to full sentences in BrocantoJ, in association to a corresponding visual animation exemplifying the corresponding move in the game, but did not actively play the game. The passive exposure set consisted of a total of 144 distinct aural stimuli (full BrocantoJ sentences) presented in association with the

corresponding game moves in 6 blocks (24 stimuli each; the full set of stimuli is provided in Appendix B). The participants were exposed to Block 1 in Session 1, to Block 2 and 3 in Session 2 and to Block 3, 4 and 5 in Session 3 according to the diagram in Figure 6.1.

In terms of the directions given to the participants, immediately after the vocabulary training and testing the children were told that whilst up to now they had learnt single words in the new language, in what followed they would hear more words spoken together and that these words would describe the game moves shown on screen. They were told that paying attention would help them make more points in the game they would play later on, and that the video would last about 4 minutes with no breaks. After each language training block the children played a game block (cf., Figure 6.1). Unlike previous studies deploying the paradigm (but similarly to Pili- Moss, 2017), the language training was administered in six short blocks instead of a single block in order to minimize the risk of drops in the children's attention.

Item frequency was counter-balanced across elements belonging to the same lexical category, although the set contained a larger number of accusative markers (ru) compared to nominative markers (ri), given that the number of OV sentences

Table 6.3a

Frequency of Vocabulary Items in the Passive Exposure Set Vocabulary items Category Frequency

blomi Noun 57 nipo Noun 57 pleca Noun 57 vode Noun 57 trose Adjective 38 neimo Adjective 38 klino Verb 12 nima Verb 44 yabe Verb 44 prazi Verb 44 noika Adverb 32 zeima Adverb 32 ri Preposition 96 ru Preposition 132

The order of presentation of the sentence stimuli was the same for all

participants. Overall, sentence length varied between 3 and 6 words in blocks 1 and 2 and between 3 and 7 words from block 3 onwards. The sentence length was kept comparatively shorter in the first part of the passive exposure because it was hypothesized that, initially, a greater amount of attentional and working memory resources would be allocated to familiarization with the mechanics of the game and the establishment/re-enforcement of initial vocabulary correspondences between aural and visual stimuli. In terms of the type of word order, each block included 14 SOV sentences, 8 OV sentences and 2 SV sentences. The frequency of SOV sentences was higher compared to the OV sentences in order to provide a sufficient number of exemplars where the linking rules between thematic interpretation and morphosyntax were expressed both syntactically (subjects preceding objects) and morphologically (subjects marked with nominative case and objects marked with accusative case). Since in OV sentences linking rules are expressed morphologically (through accusative case marking) but not syntactically, it was hypothesized that they would

have been harder to interpret. Finally, the two SV sentences were included in the set to mirror the structure of the game set (as explained in 6.3.7 the sequence of game

constellations in the game set could not be modified to exclude SV intransitive sentences). Although, overall, the number of transitive verb items was counter- balanced (as in the original BROCANTO2 studies), their distribution relative to the number of sentence stimuli assigned to each verb was skewed in individual blocks (cf., Table 6.3b).

Table 6.3b

Frequency of Verbs per Sentence Type in the Passive Exposure Set symmetric asymmetric

nima prazi yabe klino TOT

Block 1 SOV 10 2 2 - 14 OV 4 2 2 - 8 SV - - - 2 2 Tot 14 4 4 2 24 Block 2 SOV 2 2 10 - 14 OV 2 2 4 - 8 SV - - - 2 2 Tot 4 4 14 2 24 Block 3 SOV 2 10 2 - 14 OV 2 4 2 - 8 SV - - - 2 2 Tot 4 14 4 2 24 Block 4 SOV 10 2 2 - 14 OV 4 2 2 - 8 SV - - - 2 2 Tot 14 4 4 2 24 Block 5 SOV 2 2 10 - 14 OV 2 2 4 - 8 SV - - - 2 2 Tot 4 4 14 2 24 Block 6 SOV 2 10 2 - 14 OV 2 4 2 - 8 SV - - - 2 2 Tot 4 14 4 2 24 TOT 44 44 44 12 144

Investigating the effect of input for L2 learning was beyond the scope of the present study. However, the methodological choice of skewing the distribution of verbs in the passive exposure set was taken based on evidence that a skewed distribution in the frequency of lexical verbs in the input has been shown to be beneficial for the learning of linguistic regularities such as novel word order patterns (cf. Goldberg et al., 2004 and references in Chapter 4). Unlike Pili-Moss (2017), where OV sentences were introduced half way through the language training (Block 3), here all word order types were introduced from the start and occurred with constant frequency across blocks.

Each block of stimuli was created using a power-point document subsequently saved in mp4 video format. Each video lasted about 4 minutes. The Suzy character introduced the task, appeared again in a break after 16 stimuli to encourage the participants to pay attention for a little longer, and finally at the end of the video to remind the children that the next activity would be playing with the computer game (a full Suzy's script is provided in Appendix A). At the start of each block, immediately after Suzy's introduction, three consecutive slides were added to provide a countdown (3, 2, 1), with the purpose of focusing the child's attention on the start of the first stimulus slide.

Each move was exemplified by one token (in the case of the intransitive construction with klino) or two tokens (all other moves). No distractor tokens were added to the constellations in the passive exposure. Eight sectors of 4 squares each were identified on the game board and the positions the moves appeared on the board were randomized across sectors (same sequence for all participants). Although the aural sentence stimuli and move position varied for each slide, variation in the context was limited by repeating the same move for 4 consecutive slides.

Suzy's instructions and BrocantoJ words were recorded by a female Italian native speaker using monotone intonation and digitized with Audacity version 2.1.0 at a rate of 44100 Hz. Bisyllabic words were standardized in audio files lasting on average 707 ms (SD = 80.5 ms), whilst the monosyllabic words lasted on average 538 ms (SD = 4.5 ms). Each sentence stimulus was created by concatenating individual word files using AudioJoiner. Fifty milliseconds of silence were also added at the beginning and at the end of each word track for bisyllabic words and at the end of each item for monosyllabic words. Consequently there were 100 ms of silence between any two words in the final concatenated sentence stimulus, except for the transition between nouns and case markers, where the silence amounted to 50 ms. 6.3.7 Game practice

The six game blocks used in the experiment were selected from a larger set of 18 blocks used in the comprehension phase of the game in a previous BROCANTO2 study (Grey, 2014; the original blocks were numbered 1, 2, 5, 9, 10 and 13). As the paradigm had never been used with children before, the rationale behind the selection of blocks in the piloting phase was to simplify the game preferring blocks that (a) had a lower number of distractor tokens per game constellation, and (b) had the lowest possible number of moves were explicit use of adverbs was required to univocally identify the move. Using the same game configurations (which could not be modified), new BrocantoJ sentence stimuli were created for the six blocks and uploaded to the program’s XML files.

In total the game practice set consisted of 120 BrocantoJ stimuli (20 stimuli per block). Block 1 was administered in Session 1, Block 2 and 3 were administered in Session 2 and Blocks 4, 5 and 6 were administered in Session 3 (cf., Figure 6.1). The order of presentation of the sentence stimuli was the same for all participants.

Overall, sentence length varied between 3 and 8 words. Each block included 12 SOV sentences, 6 OV sentences and 2 SV sentences. To facilitate stimulus retention in short-term memory, the choice was made to keep the aural sentence stimulus as short as possible so that each sentence included only the vocabulary items necessary to univocally identify the correct move, at the expense of perfect item counter-balancing for adjectives, adverbs and token names in the game practice set (Table 6.4a; see Appendix C for the complete set). Again, the practice exposure set contained a larger number of accusative markers (ru) compared to nominative markers (ri), given that the number of OV sentences in the set exceeded the number of SV sentences.

Table 6.4a

Frequency of Vocabulary Items in the Game Practice Set Vocabulary items Category Frequency

blomi Noun 44 nipo Noun 52 pleca Noun 45 vode Noun 51 trose Adjective 58 neimo Adjective 51 klino Verb 12 nima Verb 36 yabe Verb 36 prazi Verb 36 noika Adverb 19 zeima Adverb 18 ri Preposition 84 ru Preposition 108

The distribution per block relative to the type and number of sentence stimuli assigned to each verb is reported in Table 6.4b. Unlike the set reported in Table 6.3b, creating a skewed distribution in each of the blocks was not possible in this case due to the fact that the sequence of game constellations in individual game blocks could not be modified.

Table 6.4b

Frequency of Verbs per Sentence Type in the Game Practice Set symmetric asymmetric

nima prazi yabe klino TOT

Block 1 SOV 4 4 4 - 12 OV 2 2 2 - 6 SV - - - 2 2 Tot 6 6 6 2 20 Block 2 SOV 4 4 4 - 12 OV 2 2 2 - 6 SV - - - 2 2 Tot 6 6 6 2 20 Block 3 SOV 4 4 4 - 12 OV 2 2 2 - 6 SV - - - 2 2 Tot 6 6 6 2 20 Block 4 SOV 3 6 3 - 12 OV 3 - 3 - 6 SV - - - 2 2 Tot 6 6 6 2 20 Block 5 SOV 5 4 3 - 12 OV 1 2 3 - 6 SV - - - 2 2 Tot 6 6 6 2 20 Block 6 SOV 4 5 3 - 12 OV 2 1 3 - 6 SV - - - 2 2 Tot 6 6 6 2 20 TOT 36 36 36 12 120

The word tracks for the sentence stimuli were the same as those deployed for the passive exposure set, but an additional 100 ms of silence was added between any two words in the final sentence stimulus, so that the speech rate was slower in the game practice than in the passive exposure. Prior to the start of the first game block the children were given the instruction to listen well to the words in the new language, and then make the move they thought the words were describing as fast as they could. After making their move, they were immediately given feedback on screen in the form of the words 'correct' or 'incorrect' (but were not shown the correct move in case of error). The next stimulus was presented immediately afterwards or after 60 seconds in

case of no response. A percentage correct score appeared on screen at the end of each block (corresponding to an increase of 5% for each correctly performed move, with the score starting at 0%). The participants were told that, initially, making many points was not easy, but that if they kept listening carefully they would become better and better at the game.

Online Measure of Language Learning (Overall Comprehension). During the computer game and unbeknownst to the participants the computer program created a by-trial online record of their moves and running score. As accurate performance in the game depended on the comprehension of BrocantoJ sentence stimuli, the running score provided an online overall measure of accuracy of language comprehension at item level.

Online Measure of Thematic Linking. In symmetric contexts (sentences with nima and prazi) the initial constellation does not provide cues to predict which of the two tokens will move/move first and accuracy exclusively depends on the ability to assign the correct thematic interpretation to the nouns (cf. 5.2.3). For this reason, accuracy in the subset of stimuli relative to symmetric contexts can be used to measure the learning of thematic linking. It is important to recall that although the analysis of SOV sentences in itself can provide evidence of the learning of linking rules, this is not sufficient to allow a differentiation between syntactic and

morphological linking. This is because in SOV contexts the information about thematic linking can be conveyed by the word order (which is fixed) and/or by the case markers. On the other hand, OV sentences provide the possibility to isolate morphological linking, because in these sentences correct thematic linking exclusively depends on the accurate interpretation of the accusative case marker.