Abstract:

Understanding the cognitive mechanisms underlying word segmentation has been an important focus for research on Chinese reading. Previous research shows that Chinese readers can identify words very efficiently even though the Chinese script does not use spaces between words or other visual cues to demarcate word boundaries. One possibility is that Chinese readers can make use of their knowledge of the positional frequency of characters (i.e., how often they are used as the first or second character in words) to facilitate the word segmentation and recognition. Nevertheless, this potential role for position frequency remains controversial. Accordingly, with the present research we used the lexical decision task and eye movements during reading to investigate the use of positional frequency information about the first and second character of two-character words in both visual word recognition and reading.

Four experiments were conducted to explore the effect of character positional frequency. In Experiments 1a and 1b, we manipulated both first and second character positional frequency (high vs. low) for target words that were of either high or low lexical frequency. Four types of two-character words were selected: (1) H-H: character with high word-beginning and high word-ending positional frequency. (2) H-L: character with high word-beginning and low word-ending positional frequency. (3) L-H: character with low word-beginning and high word-ending positional frequency. (4) L-L: character with low word-beginning and low word-ending positional frequency. Sixty college students were instructed to perform visual lexical decision tasks on words and pseudo-words. The results of response accuracy and latency of correct responses were reported in Table 1 (Experiment 1a) and Table 2 (Experiment 1b). In Experiment 1a, the main effects of first and second character positional frequency were not significant for response accuracy (first character, b = 0.45, SE = 1.29, z = 0.35, p = 0.73, 95% CI = [−2.07, 2.97]; second character, b = 1.38, SE = 1.29, z = 1.07, p = 0.29, 95% CI = [−1.15, 3.19]) and reaction times (first character, b = −0.01, SE = 0.02, t = −0.51, p = 0.62, 95% CI = [−0.04, 0.02]; second character, b = 0.02, SE = 0.01, t = 1.19, p = 0.24, 95% CI = [−0.01, 0.05]). The interaction between first and second character positional frequency was not observed (ps > 0.05). In Experiment 1b, the main effect of first character positional frequency was significant for response accuracy (b = 1.08, SE = 0.09, z = 11.58, p < 0.001, 95% CI = [1.26]) and reaction times (b = 0.10, SE = 0.02, t = 5.22, p < 0.001, 95% CI = [0.13]). Error rates and reaction times were less on high first character positional frequency than those words with low first character positional frequency, regardless of the second character positional frequency conditions. There was no significant main effect of second character positional frequency for response accuracy (b = 0.15, SE = 0.09, z = 1.57, p = 0.11, 95% CI = [−0.04, 0.33]) and reaction times (b = 0.00, SE = 0.02, t = −0.03, p = 0.97, 95% CI = [−0.05, 0.04]), nor interaction between first and second character positional frequency (ps > 0.05). Consequently, this aspect of our findings showed a bias for first character positional frequency to facilitate the lexical recognition.

In Experiment 2a and 2b, we used measures of eye movements to investigate the use of character positional frequency in natural sentence reading. Sixty college students were required to read sentences while their eye movements were recorded using an EyeLink 1000 eye-tracker (sampling rate = 1000 Hz). Mean eye movement measures were reported in Table 3 (Experiment 2a) and Table 4 (Experiment 2b). In Experiment 2a, the main effect of first character positional frequency was not significant for reading times (FFD, b = −0.01, SE = 0.02, t = −0.14, p = 0.89, 95% CI = [−0.04, 0.04]; GD, b = −0.01, SE = 0.03, t = −0.39, p = 0.70, 95% CI = [−0.06, 0.04]; RPD, b = 0.00, SE = 0.04, t = 0.06, p = 0.95, 95% CI = [−0.06, 0.07]; TRT, b =0.03, SE = 0.04, t = 0.67, p = 0.51, 95% CI = [−0.05, 0.11]). The main effect of second character positional frequency was not significant for reading times (FFD, b = 0.00, SE = 0.02, t = 0.21, p = 0.84, 95% CI = [−0.04, 0.05]; GD, b =0.00, SE = 0.03, t = −0.17, p = 0.86, 95% CI = [−0.06, 0.05]; RPD, b = 0.00, SE = 0.04, t = 0.02, p = 0.99, 95% CI = [−0.07, 0.07]; TRT, b = 0.02, SE = 0.04, t = 0.69, p = 0.50, 95% CI = [−0.06, 0.11]). In Experiments 2b, there was main effect of first character positional frequency on fixation durations (GD, b = 0.05, SE = 0.02, t = 2.01, p = 0.05, 95% CI = [0.10]; RPD, b = 0.075, SE = 0.03, t = 2.65, p = 0.01, 95% CI = [−0.06, 0.06]; TRT, b = 0.08, SE = 0.03, t = 3.03, p = 0.004, 95% CI = [0.13]). Reading times were shorter on high first character positional frequency than those words with low first character positional frequency, regardless of the second character positional frequency condition. The main effect of second character positional frequency was negligible (FFD, b = −0.01, SE = 0.02, t = −0.39, p = 0.69, 95% CI = [−0.04, 0.03]; GD, b = −0.02, SE = 0.03, t = −0.60, p = 0.55, 95% CI = [−0.08, 0.04]; RPD, b = 0.01, SE = 0.03, t = 0.01, p = 0.99, 95% CI = [−0.06, 0.06]; TRT, b = −0.02, SE = 0.03, t = −0.49, p = 0.63, 95% CI = [−0.08, 0.05). There were no significant interaction between first and second character positional frequency for either high or low frequency target words (ps > 0.05). The results clearly show that second character positional frequency did not influence fixation durations, while a benefit to reading times for first character positional frequency only.

The results of Experiments 1a and 1b revealed an effect of first character positional frequency in both the accuracy and latency of lexical decisions only for target words with low lexical frequency, and no effect of second character positional frequency regardless of the lexical frequency of the target word. The results of Experiment 2a and 2b showed a clear effect of first character position frequency on reading times (gaze duration, regression path reading time, and total reading time) for target words of low lexical frequency only, and no effect of second character positional frequency for either high or low frequency target words. Participants made shorter reaction times and fixation duration in high character positional frequency condition than in low character positional frequency condition for target words with low lexical frequency.

In sum, findings from both the lexical decision task and measures of eye movements in reading reveal a privileged role for first character positional frequency as a cue to word segmentation and recognition. However, this influence of character positional frequency depends on word frequency, and is only observed for words of lower lexical frequency. We argue that these findings support the augmented addressed morphology model theory of Chinese word recognition, and can help inform the development of a model of character positional processing model in Chinese reading.

Key words: word segmentation, character positional frequency, Chinese reading, eye movements