Materials

國

立 政 治 大 學

‧

N a

tio na

l C h engchi U ni ve rs it y

black color on light gray background and the reading direction of the sentences was horizontal from left to right. The characters were displayed in kai font type in a size of 32² pixels with a space of 4 pixels between each character; in this way, at a viewing distance of 70cm, one character with the space before it subtended approximately 1° of VA.

Additionally, a chinrest was used to minimize the head movements and a response console with left and right keys served as the interface for answering the comprehension questions as well as to terminate the trials.

2.2.3 Materials

In order to assess the VCD influence on saccade targeting mechanism the words’

VCD index was computed for the 2-character words of a Chinese Corpus (Academia Sinica Taiwan, 2003); afterwards, some cut points were used for selecting 3 distinctive groups of target words as follows: words with a VCD index less than -10 were considered to have a Left Bias (LB) of its inner visual components, words whose VCD was between -0.09 and 0.09 were considered unbiased or Non Bias (NB) and words with a VCD index larger than 10, were considered to have Right Bias (RB). After forming these groups, the amount of word candidates was subsequently reduced to 42 for each condition by employing a statistical-control approach on several linguistic and visual variables (i.e. word frequency, character frequency of the 1^st and 2^nd character, word stroke number and neighborhood size of the 1^st and 2^nd character). The procedure described above resulted in 3 well controlled groups of target words which only differed in their VCD index and incidentally in the 1^st and 2^nd character’s stroke number (see Table 6).

‧ 國

立 政 治 大 學

‧

N a

tio na

l C h engchi U ni ve rs it y

Table 6. Mean of word traits for Target, Pre and Postarget words

Left Bias Non Bias Right Bias F (2,123)

VCD Index -12.80 (0.21) -0.01 (0.01) 13.16 (0.36) 2909 ^**

Word Freq.log10 0.33 (0.09) 0.42 (0.07) 0.40 (0.07) 0.35 Word strokes 22.69 (0.75) 22.43 (0.26) 22.81 (0.79) 0.09 1^st character Strokes 16.62 (0.58) 11.05 (0.25) 5.62 (0.33) 177.65 ^**

2^nd character Strokes 6.07 (0.31) 11.38 (0.25) 17.19 (0.65) 158.33 ^**

1^st character Freq.log10 2.29 (0.08) 2.34 (0.09) 2.19 (0.13) 0.59 2^nd character Freq.log10 2.29 (0.09) 2.25 (0.11) 2.14 (0.11) 0.58 1^st character NS 20.26 (1.68) 20.83 (2.83) 19.71 (2.87) 0.05 2^nd character NS 21.36 (3.20) 20.59 (2.86) 21.07 (4.50) 0.01

Note: The standard error of the mean is reported in parentheses. Significance codes: ^** p<0.01; ^* p<0.05

A sentence was constructed for each target word in which target words always occupied a central position avoiding the first and the last five character positions; in addition, the words at the right and left of the target word (i.e. Pretarget and postarget words) were purposively selected to be a) 2-character words, excluding those that belonged to larger arrangements, and b) had a VCD Index around 0 (i.e. resembling in their VCD to the NB target group). The later aspect was evaluated by comparing the pretarget words by condition with the NB target words in two ANOVAs, once for the total word stroke count (F (3,164) = 1.56; p > 0.3) and once for the VCD index (F (3,164) = 1.15; p > 0.4).

‧ 國

立 政 治 大 學

‧

N a

tio na

l C h engchi U ni ve rs it y

Table 7. Mean of word traits for Target, Pre and Postarget words

Left Bias Non Bias Right Bias F (2,123)

Pretarget words

VCD Index -0.05 (0.51) 0.21 (0.64) -0.97 (0.53) 1.21 Word Freq.log10 1.87 (0.15) 1.41 (0.14) 1.77 (0.14) 2.99 Word strokes 20.19 (1.32) 20.57 (1.18) 19.45 (0.96) 0.24 1^st character Strokes 10.14 (0.76) 10.02 (0.68) 9.74 (0.52) 0.10 2^nd character Strokes 10.05 (0.69) 10.55 (0.71) 9.71 (0.54) 0.42 1^st character Freq.log10 2.85 (0.10) 2.68 (0.11) 2.91 (0.07) 1.54 2^nd character Freq.log10 2.95 (0.10) 2.84 (0.10) 3.06 (0.09) 1.19 1^st character NS 34.12 (4.58) 32.21 (4.41) 35.29 (4.86) 0.11 2^nd character NS 41.48 (4.83) 34.02 (4.51) 41.95 (4.85) 0.88

Postarget words

VCD Index -0.35 (0.59) -0.62 (0.51) 0.86 (0.60) 1.92 Word Freq.log10 1.68 (0.16) 1.22 (0.13) 1.50 (0.13) 2.74 Word strokes 19.95 (1.10) 22.98 (1.27) 20.8 (0.94) 1.95 1^st character Strokes 9.88 (0.69) 11.36 (0.67) 9.86 (0.46) 1.95 2^nd character Strokes 10.07 (0.54) 11.62 (0.73) 10.98 (0.62) 1.51 1^st character Freq.log10 2.87 (0.09) 2.66 (0.09) 2.78 (0.10) 1.18 2^nd character Freq.log10 3.05 (0.10) 2.68 (0.11) 2.59 (0.09) 5.72 ^**

1^st character NS 30.7 (4.36) 28.2 (3.49) 26.5 (3.14) 0.32 2^nd character NS 51.7 (6.83) 37.7 (5.54) 27.4 (3.89) 4.84 ^**

Note: The standard error of the mean is reported in parentheses. Significance codes: ^** p<0.01; ^* p<0.05

‧

0.2) neither for the VCD index (F (3,164) = 1.71; p > 0.2); ensuring that both, the pretarget and postarget words had mean VCD values around 0. Moreover most of the pretarget words properties were successfully controlled across conditions as a statistical control in order to reduce the influence of the immediately previous fixation on the ILP responses within target words (see Table 7).

In addition, the predictability of the target words was estimated with a clozing task²³; this procedure confirmed that all of the target words within the reading material are low predictable and do not strongly predict other words, thus reducing the context influence on saccade targeting mechanism. In this way, the sentences are not biased towards different words neither towards the target words, making the context neutral and no differences were found regarding the words’ predictability between conditions (F(2,123) = 1.29; p > 0.3).

Finally, it is worth mentioning that the syntactic category of target words was counterbalanced within each condition, half of the target words were verbs (i.e. 21) and the other half were nouns (i.e. 21).

Sentences were 20 to 27 characters in length, counting the comma symbol as one character (M= 25.34, SD=1.78), and comprised 3193 tokens of 1839 words; sentences also varied in number of words from 10 to 20 words (M=15, SD=1.78) and the contrasts

23 The predictability of the target words was rated by asking a group of Chinese native speakers different to the participants in this experiment (13 males and 7 females; mean age= 21.9, SD= 2.1) to provide a word after reading the sentence up to the target word; those sentences for which the predictability of the target word reached a ratio ≥ 0.3 (i.e. 6 out of 20 answer) were considered to be highly predictable, and therefore were excluded.

‧ 國

立 政 治 大 學

‧

N a

tio na

l C h engchi U ni ve rs it y

between conditions didn’t reached significant levels, neither by number of characters (F(2,123) = 0.66; p > 0.4) nor by number of words (F(2,123) = 0.08; p > 0.4). Some example sentences are displayed in Figure 12, additional descriptive statistics of the reading material by VCD condition and syntactic category are available in Appendix C.

Figure 12. Example sentences of the VCD conditions

Note: In blue ink the target words. The English translation for the sentences are, Left Bias: “The Chinese New Year festivities include the Lunar New Year’s Eve as well as the Spring Festival, the Lantern Festival among other activities”; Non Bias: “The peaceful church was the place I lingered the most when I was studying at Yale University”; Right Bias: “At the moment of start shooting the film the leading actor selection was unsettled and the funds haven’t been raised”.