CHAPTER 3 HOMOPHONE AND PHONOLOGICAL PRIMING OF OBJECT NAMING IN
3. Comparison of the results of Experiments 3 and 4
The repetition priming effects were similar from reading aloud words in Experiment 3 and when the words are read silently in Experiment 4 (106 vs 130ms). In order to evaluate possible differences in priming from reading aloud compared to silently, another linear mixed effect model was constructed with same fixed effects and random effect as before, but also including experiment (reading aloud vs. silent read) and all interactions. The best- fitting model included the fixed effect of prime type, prime word frequency, picture naming frequency and the interaction between experiment and priming word frequency. Other interactions were found to be non-significant and were therefore removed from the model. Summary statistics for the model are shown in Table 3.6 and Figures 3.5 and 3.6 show the magnitudes of the priming effects observed in the two experiments.
However, it must be accepted that, as the two studies have n=28 and n=24 participants, this between-group contrast is statistically underpowered, which limits the conclusions that may be drawn from this contrast.
Figure 3.2 The Homophone Effects (With Standard Errors) in Two Experiments
Figure 3.3 The Homophone Effects (With Standard Errors) in Two Experiments
Higher frequency Same frequency Lower frequency
Experiment 3 76 84 81 Experiment 4 86 83 73 0 20 40 60 80 100 120 Fa cilit at ion in P ic tur e n am in g (m s)
Homophone Effect
Higher frequency Same frequency Lower frequency
Experiment 3 24 55 39 Experiment 4 53 59 44 0 10 20 30 40 50 60 70 80 Fa ci li ta ti on in P ic tu re n am in g (m s)
Phonological Effect
Table 3.6 Summary Statistics of the Mixed-Effect Model of Picture Naming Latencies (N=10,819).
Fixed Part B S.E. z-ratio p-value Intercept 758.56 9.15 82.87 <.001
Phonological Relatedness = Same syllable -51.14 4.69 -10.91 <.001
Phonological Relatedness = Homophonic -80.15 6.84 -11.72 <.001
Priming word frequency -10.05 1.65 -6.11 <.001
Picture name frequency -25.33 3.84 -6.59 <.001
Experiment condition (Silent read) ×
7.54 2.22 3.39 0.001
Priming word frequency
Random Part Variance SD
Level: Subject
Intercept 3674.72 809.56
Phonological Relatedness = Same syllable 690.07 214.61
Phonological Relatedness = Homophonic 1854.13 454.29
Priming word frequency 7.87 13591.00
Picture name frequency 0.00 0.00
Level: Picture
Intercept 2158.06 281.65
Phonological Relatedness = Same syllable 377.19 409.81
Phonological Relatedness = Homophonic 393.09 408.81
Priming word frequency 230.98 61.81
Level: Word
General Discussion
Experiments 3 and 4 examined the priming of object naming by the prior reading of a single character (and so monosyllabic) Chinese word which was the name of the object, a homophone of the object’s name, a phonetically related word to the object name (by having the same syllable but a different tone), or an unrelated word. Both experiments produced a clear and very similar pattern of results. There was a strong repetition (or identity)
priming effect of 106ms in Experiment 3 and 130ms in Experiment 4. There was a smaller but still substantial effect of homophone priming of 80ms in Experiment 3 and 40ms in Experiment 4. Finally, there was a smaller, but still reliable, effect of phonological priming of 40ms in Experiment 3 and 52ms in Experiment 4.
Whether the prime word was read aloud or read silently appeared to make very little difference to either the magnitude or the pattern of priming effects observed. The
homophone and phonetic priming effects observed do not require the explicit articulation of the prime words, and this suggests that in Chinese reading the phonology of the words becomes activated automatically, irrespective of reading task. Further, the results showed that the priming effects were not affected by the frequency of the prime words. The frequency effects observed on word reading latencies can be interpreted as reflecting the ease of access to, or retrieval from, permanent lexical-level representations in either the orthographic word recognition system or the phonological word production system, or both. The results of both experiments are more consistent with the notion that there are
independent rather than shared phonological lexical representations of homophones. The word-specific frequency will affect the ease of accessing and/or retrieving the lexical phonology of homophones (and indeed all words); but once retrieved, the persisting activation produced to influence priming of object naming appears to be unaffected by frequency.
The general interpretation of these results is that priming can result from persisting activation, produced by word reading, at various levels of processing underlying spoken word production. Repetition priming reflects persisting activation at semantic, lexical, sub- lexical, and articulatory levels. Homophone priming reflects persisting activation at lexical, sub-lexical, and articulatory levels (including syllable motor programs). Phonetic priming reflects persisting activation at sub-lexical and articulatory levels, and in terms of Roelofs’ (2015) and Zhu, Zhang, and Damian’s (2016) models of phonological encoding in Chinese speech, how segmental phonemes within a monosyllabic word are integrated with the word’s tonal frame to produce syllable motor programs. Homophone prime words activate the same syllable motor patterns as required to produce the target object names, whereas phonetically related prime words (that have the same phonetic syllable but not the same tone) will not. In the two experiments, only single syllable words were used. To test Roelofs’ theory further, it might be interesting (if practically very difficult) to attempt to dissociate possible effects from atonal syllables (and their phoneme segments) and tonal frames. This would involve selecting probably multi-syllable words with the same syllables but different tones and words with the same tonal frames but different syllables. However, the possibility of detecting “tone-only” priming is limited due to the fact that there are only four tones used in Mandarin Chinese (although Cantonese has more tones).
Chinese reading is based primarily on the connections between orthography and meaning. This has led to the widespread assumption that skilled Chinese readers tend to rely only upon orthographic and semantic information when processing visually presented characters (Sze, Yap, & Rickard Liow, 2015). However, the results from the experiments presented here, along with others, suggest that phonological information is activated internally in the processing of Chinese characters. The results of Experiment 4 further extend this notion by showing that effects of a character’s phonology can occur even in silent reading.
In a task in which participants were asked to judge if two words were semantically related, Xu, Pollatsek, and Potter (1999) found that phonological interference was
observed only in exact homophones and not in characters that have same consonant and vowel. In Experiments 3 and 4, it is possible that the prime word activated a range of semantically related words, and also possible that the phonology of these related words were activated. However, in order for participants to read the character aloud, the presentation time of the prime word was set at 1000ms for both the reading aloud and silent reading conditions. One might argue that during this presentation time (and the 500ms of inter-stimulus interval), the participants had ample time to both finish the processing of the character and to fully activate the phonology of likely targets (and in the experiments there was a 4 in 10 chance that the target object name would be
phonologically identical to the phonology of the prime word). Possibilities along these lines could be tested in future research by manipulating both the exposure duration of the prime word and, by the inclusion of semantically unrelated or related filler words, the probability of phonological match between prime word and target object name.