Parallel morpho-orthographic and morpho-semantic activation in processing second language morphologically complex words: Evidence from Chinese-English bilinguals

Abstract

Objectives/research questions:

The present study focused on the performance of Chinese learners of English to investigate the activation of morpho-semantic information in the early processing of second language (L2) morphologically complex words when participants’ first language (L1) and L2 are typologically different.

Methodology:

We used forward masked priming paradigm to compare the priming effect in three prime conditions, semantically transparent, semantically opaque and semantically related. In Experiment 1, the stimulus onset asynchrony (SOA) was 40 milliseconds (ms), while, In Experiment 2, the SOA was extended to 80 ms.

Data and analysis:

Reaction time and comprehension accuracy data were analysed using the linear mixed-effects model.

Findings/conclusions:

In Experiment 1, we only found a priming effect in the semantically transparent condition. In Experiment 2, we found a reliable priming effect in the semantically opaque condition, but not in the semantically related condition. These results suggest even Chinese learners of English whose native language is typologically different from English can employ the rule-based decomposing mechanism. The decomposition is based on the interplay of morpho-orthographic and morpho-semantic information, adding new evidence to the assumption of parallel orthography-semantics activation.

Originality:

We manipulated SOAs to investigate the role of the morpho-semantic factor in the early processing of morphologically complex L2 words when participants’ L1 and L2 are typologically different.

Keywords

Second language morphologically complex words morpho-orthographic morpho-semantic priming

Introduction

A morpheme is the minimal functional unit composing a word, and the minimal meaning-bearing unit (Raveh & Rueckl, 2000). In a natural language, most of the words are morphologically complex words, which are formed by stems and affixes, such as ‘management’, which is formed by the stem ‘manage’ and the affix ‘ment’. Marslen-Wilson (2007) concluded that the decomposition of morphologically complex word is ‘one of the highest priorities’ (p. 189) of the system. Researches on morphologically complex words could help us have a better understanding of word processing and language comprehension.

Many studies have found morphological decomposition in the early processing of morphologically complex words (Beyersmann, Castles, & Coltheart, 2011; Diependaele Sandra, & Grainger, 2005; Diependaele, Duñabeitia, Morris, & Keuleers, 2011; Marslen-Wilson, Tyler, Waksler, & Older, 1994; McCutchen, Logan, & Biangardi-Orpe, 2009; Neubauer & Clahsen, 2009; Raveh & Rueckl, 2000; Silva & Clahsen, 2008; Sonnenstuhl, Eisenbeiss, & Clahsen, 1999; Taft, 1981; Taft & Forster, 1975). However, it is not clear about the sequence of morpho-orthographic and morpho-semantic activation during this process. Generally, there are two different accounts for the above question, the form-then-meaning account and the form-with-meaning account.

The form-then-meaning account assumes that a word’s orthographic form must be processed before its meaning becomes available. Researchers typically assume that the later stages of word processing are influenced by semantic properties of the stem but the initial stages are solely orthographic (Beyersmann et al., 2011; Rastle, Davis, & New, 2004; Rastle & Davis, 2008; Rueckl & Aicher, 2008). Contrarily, according to the form-with-meaning account, morpho-orthographic and morpho-semantic information should be activated at the same time during the early stages of word processing, and they both affect the early processing of morphologically complex words (Feldman & Soltano, 1999; Feldman, Soltano, Pastizzo, & Francis, 2004; Feldman, O’Connor, & del Prado Martín, 2009; Giraudo & Grainger, 2001; Marelli & Luzzatti, 2012).

Most of the studies looked into morpho-orthographic and morpho-semantic activation in visual word recognition using the masked priming paradigm, by comparing the prime effect of the semantically transparent word with that of the semantically opaque word. Here, semantically transparent word refers to the word that bears similar meaning with its stem (e.g. worker-work); semantically opaque word refers to the word that has a different meaning from its stem (e.g. department-depart), or the pseudo-complex word that has false morphological status, even though it comprises the letter pattern of a stem and an affix (e.g. corner-corn). Usually, the primes are presented for around 30–80 ms after the mask (####) and then the target words follow. After the target words disappear, participants are required to accomplish a lexical decision task or a word naming task. Because of the short duration of the primes, participants are unaware of them, and thus it is possible for researchers to look into the early stages of word processing through this paradigm (Silva & Clahsen, 2008).

The form-then-meaning account and the form-with-meaning account make different predictions. If the priming effects of semantically transparent words and semantically opaque words are similar in magnitude, this means the morpho-semantic information is irrelevant to the early processing of morphologically complex words, and the early stages of processing are based on the morpho-orthography. If the priming effects of semantically transparent words are greater than semantically opaque words, or the former occurs earlier than the latter, this means morpho-semantic information does affect the early processing of morphologically complex words, and both morpho-orthographic and morpho-semantic information are activated in the early stages of word processing.

Using the masked priming paradigm with 42 ms stimulus onset asynchrony (SOA), Rastle et al. (2004) looked into the priming effect in three prime conditions: semantically transparent prime, semantically opaque prime and orthographically related prime (e.g. brothel-broth, in which -el is not real affix). They found a similar priming effect in the semantically transparent and semantically opaque conditions, which was greater than the priming effect in the orthographically related condition. Since there was no difference in the priming effect between the semantically transparent and semantically opaque conditions, their results support the form-then-meaning account that only morpho-orthographic information drives the early processing of words. Some other studies also found both semantic transparency and semantic opaqueness facilitate the early processing of the stem (Boudelaa & Marslen-Wilson, 2005; Kazanina, Dukova-Zheleva, Geber, Kharlamov, & Tonciulescu, 2008; Longtin, Segui, & Halle, 2003; Marslen-Wilson, Bozic, & Randall, 2008; Rastle & Davis, 2003, 2008).

However, other researchers got different results using the same masked priming paradigm. For example, Diependaele, Sandra, and Grainger (2005, 2009) found that the priming effect of semantically transparent words occurred earlier than that of semantically opaque words, indicating that morpho-semantic information affects the early processing of morphologically complex words. Morris, Frank, Grainger, and Holcomb (2007) examined the role of semantics in the segmentation of morphologically complex words by examining Event-Related Potentials (ERPs) to targets primed by semantically transparent, semantically opaque and orthographically related words using the masked priming paradigm. They found both N250 and the N400 effects with transparent items generating greater priming than the orthographically related or opaque ones, suggesting that morpho-semantic information is activated at the early stages of processing morphologically complex words. Feldman et al. (2009) found that semantically transparent primes show a larger priming effect than opaque or pseudo-complex primes. In their opinion, both morpho-orthographic and morpho-semantic information may affect the early processing of morphologically complex words.

The current disagreement is mainly based on evidence from native (L1) speakers, so we find it interesting to turn our eyes to second language (L2) processing. Does the morpho-semantic factor play a role at the early stages of processing L2 morphologically complex words?

Up until now, few studies have focused on the morpho-semantic role in the early processing of L2 morphologically complex words. Most of studies look into the difference between native speakers and L2 learners in processing morphologically complex words (Kirkici & Clahsen, 2012; Neubauer & Clahsen, 2009; Silva & Clahsen, 2008). For instance, Silva and Clahsen (2008) found that native English speakers’ processing of regular inflected and derivative words is based on decomposition, while Chinese-English learners, German-English learners and Japanese-English learners rely on the whole word. Although the latter do decompose a little when processing derivative words, their decomposition was weaker than native English speakers. Neubauer and Clahsen (2009) compared German native speakers and participants who use German as a L2, and they found that native German speakers processed regular inflected words in a decomposing way, while L2 German learners treated them as whole words. This result was further confirmed by Kirkici and Clahsen (2012). However, other studies found that L2 learners are capable of decomposition, just like native speakers. For example, Pliatsikas and Marinis (2013) compared highly proficient Greek-English learners to native English speakers on a self-paced reading task involving past tense forms embedded in plausible sentences. Results showed rule-based decomposition in L2 learners. Liang and Chen (2014) investigated the effect of L2 proficiency on processing L2 morphologically complex words in Chinese learners of English. ERP results showed that highly proficient L2 learners demonstrated the priming effect within 350–550 ms in the morphological condition, associating with an N400 reduction, while less proficient L2 learners showed no morphological priming effect within the N400 range. The ERP results indicate that highly proficient L2 learners manifest rule-based decomposition. Taken all together, no agreement has been reached yet on the question of whether L2 learners are capable of native-like processing of morphologically complex words.

Besides, another line of research is about the effect of L1 characteristics on processing L2 morphologically complex words and, still, no agreement has been reached yet. For example, Basnight-Brown, Chen, Shu, Kostić, and Feldman (2007) found that Serbian-English bilinguals, but not Chinese-English bilinguals, showed facilitation due to form overlap between irregular past and present tense forms with a nested stem. Serbian is an inflectional language with alphabetic script, and like English, it also belongs to the Indo-European language family. However, Chinese is a non-inflectional language, and belongs to the Sino-Tibetan language family (Chu, 1982). The Chinese graphemic unit corresponds to a whole syllable, not letters. So different L1 characteristics may be the reason why Serbian-English and Chinese-English bilinguals show different performances in Basnight-Brown et al. (2007)’s study. However, Silva and Clahsen (2008) investigated regular past tense forms and deadjectival nominalizations with -ness and -ity in adult native speakers of English and in different groups of advanced adult L2 learners of English. They found parallel priming patterns for German and Chinese L2 learners, indicating that L1 characteristics do not affect the processing of L2 morphologically complex words. Also, in Liang and Chen (2014)’s study, highly proficient Chinese learners of English decompose morphologically complex words as native speakers do, despite the differences between Chinese and English in inflection.

In a word, most previous studies cared about the different mechanisms used between native speakers and L2 learners, and the effect of L1 characteristics when they process morphologically complex words. Up until now, only one study carried out by Diependaele et al. (2011) looked into whether the morpho-semantic factor influences the early processing of L2 morphologically complex words when the SOA was 53 ms. They compared Spanish-English and Dutch-English learners to native English speakers when they process derivative words in English. They found similar priming patterns for the native English speakers and the two groups of bilinguals, with the priming effect being largest with semantically transparent primes, smallest with orthographically related primes and intermediate with semantically opaque primes. They concluded that for L2 learners both morpho-semantic and morpho-orthographic information are activated at the early stages of L2 morphologically complex word processing.

Nevertheless, it should be pointed out that the native languages of the L2 learners in Diependaele et al. (2011)’s study were Spanish and Dutch, which share some similarity with English and also belong to the Indo-European language family. Since previous studies found that L1 characteristics may affect the processing of L2 morphologically complex words (e.g. Basnight-Brown et al., 2007), it is possible that the above findings about the morpho-semantic factor in Diependaele et al.’s (2011) study would be different if a new group of L2 learners with typologically different L1 and L2, such as Chinese (L1)-English (L2), are tested. Besides, only one SOA, 53 ms, was used in this study, and we do not know whether the priming effects will be different between semantically transparent primes and semantically opaque primes with shorter or longer SOAs.

Therefore, to get a further understanding of the role of morpho-semantic factor, the present study focused on the performance of Chinese learners of English to investigate the activation of morpho-semantic information in the early processing of L2 morphologically complex words when participants’ L1 and L2 are typologically different. We have two different SOAs (Experiment 1, SOA = 40 ms; Experiment 2, SOA = 80 ms). According to the form-with-meaning account, we expect to observe an earlier priming effect in the semantically transparent condition than the semantically opaque condition or greater priming effect in the former condition than the latter. Otherwise, if there is no significant difference in the priming effect between the two conditions, we will say morpho-semantic information does not influence the early processing of morphologically complex words, supporting the form-then-meaning account.

Experiment 1

In Experiment 1 we used the masked priming paradigm and visual word recognition task to investigate whether semantically transparent words and semantically opaque words would have different priming effects when the SOA was 40 ms.

Participants

Forty students (18–25 years old, M = 22.50 ± 1.77 years, 26 females) from Beijing Normal University participated in Experiment 1. All of them were native speakers of Chinese, and they started to learn English at about 10 years old. The participants’ mean score in the College English Test Band 6 (CET-6) was 490. CET-6 is a test designed by the Ministry of Education in China, and it is used in all universities in China to evaluate the English proficiency of students who are non-English majors. It consists of tasks on listening, reading, vocabulary knowledge, grammar knowledge and writing. The total score is 710, and the cutoff point (set by the Ministry of Education) for success and failure in the test is 425. All participants were required to self-rate their listening, reading, speaking and writing skills on a five-point scale (1 for ‘very unskilled’ and 5 for ‘very skilled’). Besides, all participants took the Oxford Placement Test (OPT), which is acknowledged to be the test of English proficiency. The OPT test includes 25 multiple choice questions and a cloze test, and the total score is 50. The summary of participants’ mean scores in CET6, OPT and self-rating is presented in Table 1, from which we could tell that the participants had intermediate L2 proficiency level.

Table 1.

Mean age, second language proficiency ratings and Oxford Placement Test (OPT) scores of the participants (SD) in Experiment 1.

	Age	CET-6	OPT	Listening	Speaking	Reading	Writing
Mean score	23 (1.77)	489 (37.37)	40 (3.55)	3.39 (0.84)	3.52 (0.59)	2.87 (1.01)	3.48 (1.08)

CET-6: College English Test Band 6.

Design

We used 3 (prime type: semantically transparent, semantically opaque and semantically related) × 2 (the relatedness: related and unrelated) within-subjects design.

Material

For the three prime types – semantically transparent, semantically opaque and semantically related – there were 16 words in each condition and 48 target words in total. The target words were mono-morphological words. Semantically transparent primes refer to those that are transparent suffixed morphological relatives of the targets (e.g. reporter-report); semantically opaque primes refer to those that are pseudo-complex primes or opaque suffixed morphological relatives of the targets (e.g. corner-corn). In the transparent and opaque conditions, the overlap of letters between the primes and targets was 68.39% and 65.97%, and there was no difference between the two (t (1, 31) = 1.69, p = 0.247), indicating that the word-form similarity between primes and targets was the same in both conditions. Semantically related primes refer to those that are related to the targets semantically, but not similar orthographically, and there is no morphological connection between targets and primes (e.g. choice-selection). Unrelated primes were those that bear no orthographic, semantic and phonetic relationship with the targets (e.g. north-dress). Each prime condition has its own unrelated primes.

We matched the frequency of our materials using the SUBTLEX database (Brysbaert & New, 2009). In order to make sure that all the words were familiar to the participants, we asked another group of 30 students from Beijing Normal University who were at the same proficiency level with our participants to assess the familiarity of the materials using a five-point scale (1 for ‘don’t know’ and 5 for ‘very familiar’). Results showed that the primes and targets were all familiar words. One limitation in the materials was that the length of the target words could not be matched. So we included length as a covariate factor in the statistical analysis. The detailed information for materials is presented in Table 2.

Table 2.

The word length, frequency and familiarity of the primes and targets (SD).

	Semantically transparent		Semantically opaque		Semantically related
	Related	Unrelated	Related	Unrelated	Related	Unrelated
Primes
Length	9.13 (0.89)	9.13 (0.89)	6.81 (1.38)	6.81 (1.32)	7.00 (1.03)	7.00 (1.03)
Frequency	13 (7.52)	17 (11.12)	20 (14.35)	21 (20.71)	28 (33.78)	27 (17.05)
Familiarity	4.63 (0.25)	4.75 (0.25)	4.56 (0.40)	4.69 (0.46)	4.57 (0.29)	4.65 (0.46)
Targets
Length	6.31 (0.79)		4.50 (1.03)		5.94 (1.69)
Frequency	23 (28.91)		21 (23.68)		31 (31.73)
Familiarity	4.79 (0.21)		4.63 (0.30)		4.54 (0.27)

To make sure participants saw each target once, we split the materials into two lists, and participants only took one of them randomly. In addition, there were 48 pseudo-word targets as fillers that were made out of real words by changing one to two letters. The primes for the pseudo-word targets and real-word targets had similar frequency range and length.

Procedure

We used E-prime 1.1 software under a Windows 2000 system. The resolution ratio of the screen was 800×600, and the refresh rate was 75 Hz. Each trial began with a 500 ms fixation (+) in the middle of the screen, after which the forward mask (******) appeared and lasted for 500 ms, and then the prime was presented for 40 ms (with a 13.33 ms refresh cycle), which was followed by the target presented for 500 ms. The target was followed by a blank screen for 2000 ms. Participants were required to judge whether the target was a real word or not by pressing buttons as accurately and quickly as possible. In order to balance participants’ responses, half of our participants were asked to press the ‘J’ button if the target was a real word and ‘F’ button if the target was a pseudo-word, and the other half of the participants were asked to press the buttons the other way round. In order to reduce the amount of pure visual priming, different fonts were used for primes and targets in each prime–target pair (Clahsen & Neubauer, 2010; Heyer & Clahsen, 2015). The primes were presented in Arial, 24-font size and the targets were presented in Times New Roman, 24-font size. There were 96 trails, half of which had real words as targets while the other half had pseudo-words as targets. Before the experiment, there were practice trails for participants to get familiar with the whole procedure.

Results

Only response times (RTs) for correct responses were analysed. RTs beyond three standard deviations of the grand mean were excluded (0.48%). The mean RTs and error rate are summarized in Table 3.

Table 3.

Mean response times (RTs) (ms) and error rate (ER in %) in Experiment 1 (SD).

Prime type	Related		Unrelated		Priming effect
	RT	ER	RT	ER	Priming effect
ST	656 (173)	12 (0.09)	684 (183)	13 (0.11)	28
SO	631 (155)	12 (0.11)	637 (160)	11 (0.09)	6
SR	671 (160)	16 (0.14)	680 (156)	17 (0.14)	9

ST: semantically transparent; SO: semantically opaque; SR: semantically related.

Logarithmic transformation was conducted on the RT data and the linear mixed-effects model was used for data analysis (Baayen, Davidson, & Bates, 2008). In the model, the prime type and the relatedness were fixed factors; the word length of the target was a covariant factor; participants and items were random factors. The results showed that the main effect of prime type was not significant, F(2,78) = 1.91, p = 0.15; the main effect of relatedness was significant, F(1,39) = 7.34, p = 0.007; the main effect of word length was significant, F(1,39) = 7.41, p = 0.007; the interaction between prime type and relatedness was not significant, F(2,78) = 1.34, p = 0.26. In order to look into the priming effect of each prime type, we constructed sub-models for each type, including the fixed factor relatedness, and the random factors of participants and items. The results are summarized in Table 4.

Table 4.

Analysis of the priming effect in different prime conditions in Experiment 1.

Prime type	Value	SE	t	Pr(\|>t\|)
ST
Intercept	2.80	0.01	276.23	<0.001
Relatedness	0.02	0.006	2.61	0.009
SO
Intercept	2.79	0.009	315.07	<0.001
Relatedness	0.006	0.005	1.14	0.26
SR
Intercept	2.82	0.009	311.48	<0.001
Relatedness	0.003	0.005	0.71	0.48

ST: semantically transparent; SO: semantically opaque; SR: semantically related; SE: standard error.

From Table 4 we can see that the RT in the semantically transparent condition was significantly shorter than that in the unrelated condition, t = 2.93, p = 0.004, so the semantically transparent primes had a significant priming effect. However, the semantically opaque primes had no significant priming effect, t = 1.19, p = 0.24, and nor did the semantically related primes, t = 0.71, p = 0.48.

In analysing participants’ accuracy, we used mixed-logic regression (Baayen et al., 2008). In the model, the prime type and relatedness were fixed factors, the word length was a covariant factor and participants and items were random factors. The results showed that the main effect of prime type was not significant, F(2,78) = 2.77, p = 0.10; the main effect of relatedness was not significant, F(1,39) = 1.40, p = 0.17; the main effect of word length was not significant, F(1,39) = 0.70, p = 0.70; the interaction between prime type and relatedness was not significant, F(2,78) = 0.56, p = 0.77.

To sum up, in Experiment 1, we found a significant priming effect in the RT data in the semantically transparent condition, but not in the semantically opaque primes. This means morpho-semantic information drives the early processing of morphologically complex L2 words. To further analyse the priming effect in the semantically transparent, opaque and related conditions, we extended the SOA to 80 ms.

Experiment 2

Participants

Forty students (18–25 years old, 31 females) from Beijing Normal University participated in Experiment 2. None of them participated in Experiment 1. All participants were native speakers of Chinese, and they started to learn English at about 10 years old. The mean score of CET-6 was 492. All participants were required to self-rate their listening, reading, speaking and writing skills on the five-point scale, and took the OPT. The summary of participants’ mean scores in CET-6, OPT and self-rating is presented in Table 5. There was no significant difference between participants in Experiment 1 and 2 in their scores in CET-6, OPT and self-rating, ps > 0.1.

Table 5.

Mean age, second language proficiency ratings and Oxford Placement Test (OPT) scores of the participants (SD) in Experiment 2.

	Age	CET-6	OPT	Listening	Speaking	Reading	Writing
Mean score	22 (2.08)	492 (41.39)	39 (4.76)	3.24 (0.89)	3.35 (0.77)	3.09 (0.97)	3.32 (1.12)

CET-6: College English Test Band 6.

The design, materials and procedure were the same as Experiment 1 except that the SOA in Experiment 2 was 80 ms.

Results

Only RTs for correct responses were analysed. RTs beyond three standard deviations of the grand mean were excluded (1.18%). The mean RTs and error rate are summarized in Table 6.

Table 6.

Mean response times (RTs) (ms) and error rate (ER in %) in Experiment 2 (SD).

Prime type	Related		Unrelated		Priming effect
	RT	ER	RT	ER	Priming effect
ST	707 (222)	11 (0.09)	720 (190)	13 (0.11)	14
SO	660 (183)	8 (0.07)	684 (179)	8 (0.07)	24
SR	736 (204)	11 (0.10)	718 (183)	11 (0.10)	−18

ST: semantically transparent; SO: semantically opaque; SR: semantically related.

Logarithmic transformation was conducted on the RT data and the linear mixed-effects model was used for data analysis. In the model, the prime type and the relatedness were fixed factors, the word length of the target was a covariant factor and participants and items were random factors. The results showed that the main effect of prime type was significant, F(2,78) = 6.12, p = 0.003; the main effect of relatedness was significant, F(1,39) = 7.00, p = 0.008; the main effect of word length was marginally significant, F(1,39) = 4.05, p = 0.05; the interaction between prime type and relatedness was significant, F(2,78) = 3.92, p = 0.02. In order to look into the priming effect of each prime type, we constructed sub-models for each type, including the fixed factor relatedness, and the random factors of participants and items. The results are summarized in Table 7.

Table 7.

Analysis of the priming effect in different prime conditions in Experiment 2.

Prime type	Value	SE	t	Pr(\|>t\|)
ST
Intercept	2.83	0.013	217.59	<0.001
Relatedness	0.009	0.006	1.72	0.08
SO
Intercept	2.81	0.011	247.52	<0.001
Relatedness	0.018	0.005	3.29	0.001
SR
Intercept	2.85	0.012	238.62	<0.001
Relatedness	−0.003	0.005	−0.62	0.54

ST: semantically transparent; SO: semantically opaque; SR: semantically related; SE: standard error.

From Table 7, we can see that the priming effect of semantically transparent primes was marginally significant, t = 1.72, p = 0.08; semantically opaque primes had a significant priming effect, t = 3.29, p = 0.001, as the RTs for semantically opaque primes were significantly shorter than those in unrelated prime pairs; semantically related primes had no significant priming effect, t = −0.62, p = 0.54.

In analysing accuracy, we used mixed-logic regression. In the model, the prime type and relatedness were fixed factors, the word length was a covariant factor and participants and items were random factors. The results showed that the main effect of prime type was marginally significant, F(2,79) = 2.37, p = 0.07; the main effect of relatedness was not significant, F(1,39) = 0.09, p = 0.83; the main effect of word length was not significant, F(1,39) = 0.02, p = 0.90; the interaction between prime condition and relatedness was not significant, F(2,78) = 0.14, p = 0.86.

In Experiment 2, after we extended the SOA to 80 ms, semantically transparent primes showed a marginally significant priming effect. More importantly, semantically opaque primes showed a significant priming effect in the RT data.

Discussion

The present study investigated the role of the morpho-semantic factor in the early processing of L2 morphologically complex words. We found that only semantically transparent primes had a priming effect when the SOA was 40 ms (in Experiment 1). When the SOA was extended to 80 ms (in Experiment 2), semantically opaque primes had a significant priming effect. These findings showed that the priming effect occurred earlier in the semantically transparent condition (SOA = 40m s) than in the semantically opaque condition (SOA = 80 ms), which means at the early stages of word processing, morpho-semantic information is activated, and both morpho-orthographic and morpho-semantic factors affect the early processing of morphologically complex words at the same time. These findings support the form-with-meaning account of processing L2 morphologically complex words.

Diependaele et al. (2011) found in Spanish-English and Dutch-English speakers that when the SOA was 53 ms the priming effect was largest with transparent primes, smallest with form primes and intermediate with opaque primes. Thus, the authors conclude that the early processing of morphologically complex words is based on the interplay of morpho-orthographic and morpho-semantic information. In their study, participants’ L1 (Spanish and Dutch) and L2 (English) share some similarities and they all belong to the Indo-European language family. Since previous studies found that L1 characteristics may affect the processing of morphologically complex words (e.g. Basnight-Brown et al., 2007), the present study chose Chinese-English bilinguals, and observed the effect of the morpho-semantic factor in the early decomposition of morphologically complex words, indicating that even Chinese English learners whose native language is typologically different from English enjoy a kind of language processing mechanism that is similar to English natives. Our findings add more evidence to the assumption of parallel orthography-semantics activation in the early processing of L2 morphologically complex words.

Moreover, in the present study, the semantically transparent and opaque conditions did not differ in the overlap of letters between primes and targets, so the desynchrony found in the priming effects of the two conditions was not due to the similarity of prime–target word forms. Some previous studies also found priming effects for morphologically related primes, which could not be attributed to the form or semantic overlap between primes and targets, in visual masked priming experiments with short SOAs within 30–60 ms (Boudelaa & Marslen-Wilson, 2005; Frost, Kugler, Deutsch, & Forster, 2005; Rastle et al., 2004). In our study, semantically related primes showed no significant priming effect in both SOAs, and this further indicated that the semantically transparent priming effect could not be attributed to the semantic overlap between primes and targets.

In the present study, the length of the target words, a covariant factor, was included in the analysis, and a significant length effect was revealed showing shorter RT for words with fewer letters. Table 2 shows that the length of the targets in the semantically transparent and opaque conditions did not match; they were shorter in the opaque and longer in the transparent. Hence, could the difference between semantically transparent and opaque conditions in the priming effect be attributed to their mismatch in the length of targets? We do not think that it could. Otherwise, the results in Experiments 1 and 2 should be consistent. However, we found in Experiment 2 significant priming effect in the semantically opaque condition, which did not show up in Experiment 1. Thus, the interference of length of targets is ruled out.

We think that the priming effect found in the semantically transparent and opaque conditions is based on the analysis of the morphological structure by decomposing it into stem + affix. How does the interplay of morpho-orthography and morpho-semantics drive the processing of morphologically complex words? We would like to explain this by referring to the Mixed Model of morphological processing (Diependaele et al., 2009). According to this account, there are three levels in processing morphologically complex words. One is morpho-orthographic processing, that is, decomposing ‘farmer’ into two morphemes ‘farm’ and ‘er’; the other one is lexical form processing, that is, the facilitation of ‘farm’ to ‘farmer’ because of the letters shared in the two words; the last one is morpho-semantic processing, that is, the decomposition of ‘farmer’ into stem ‘farm’ and affix ‘er’ and the assignment of semantic meaning to the affix ‘er’ to indicate people who occupy some certain jobs related to the meaning of the stem. Besides, there are two parallel processing routes — one is based on the sub-lexical form processing, and the other is based on the meaning processing. Being different from the morpho-orthographic processing route, the morpho-semantic processing route needs to reach the lexical form first, and get to the morpho-semantics next. There are interactions between the two routes. The lexical form processing may affect the processing of morpho-semantics, while the morpho-semantic processing may also influence the lexical form processing.

Therefore, the priming effect is due to the decomposition of morphologically complex words into stem and affix. When processing semantically transparent primes, both morpho-orthographic and morpho-semantic routes are activated, and they can facilitate the processing of the target, which is the stem of the prime. When processing semantically opaque primes, only the morpho-orthographic route is activated, and thus the priming effect is weaker or later than that in the semantically transparent condition. This explains why we only found a priming effect in the semantically transparent condition when the SOA was 40 ms.

In the declarative/procedural model, Ullman (2001, 2004) claimed that there are two systems of memory: declarative memory and procedural memory. Whole-word representation relies on declarative memory while decomposition relies on procedural memory. Late L2 learners rely more on declarative memory in L2 processing because of their late acquisition age, and thus their representation for L2 words should be whole-word based (Ullman, 2001, 2004; Ullman et al., 2005). Many researchers found native speakers tend to decompose regular morphologically complex words in online processing (Lehtonen, Monahan, & Poeppel, 2011; Lück, Hahne, & Clahsen, 2006; Newman, Ullman, Pancheva, Waligura, & Neville, 2007; Silva & Clahsen, 2008), while it remains controversial whether L2 learners are capable of decomposition or not. Some researchers found that L2 learners rely on whole-word representation in processing morphologically complex L2 words. For example, Silva and Clahsen (2008) found a significant difference between advanced adult English learners and native English speakers in a masked priming lexical decision task, with L2 learners showing no morphological priming for inflected words and reduced priming for derived words. The authors concluded that L2 learners rely more on lexical storage in processing morphologically complex words. Neubauer and Clahsen (2009) obtained similar results, and thought L2 learners were less sensitive to morphological structure than native speakers.

However, some other researchers had different findings. Diependaele et al (2011) found that both late Spanish-English and late Dutch-English learners can decompose morphologically complex words. Pliatsikas and Marinis (2013) compared highly proficient Greek-English learners to native English speakers on a self-paced reading task involving past tense forms embedded in plausible sentences. Their results showed rule-based decomposition in L2 learners. Consistently, results from our present study found a priming effect in semantically transparent and opaque conditions. Based on these findings, we reckon that L2 learners cannot be constrained by the whole-word representation mechanism the whole time, and they can also develop the decomposing mechanism at a certain point. Still, this development may be modulated by L2 proficiency. Participants in the present study have been learning English for more than 10 years, and have reached an intermediate proficiency level. This may help them in decomposing morphologically complex words. Contrarily, when participants have low L2 proficiency, they may have to rely on whole-word storage.

In conclusion, the present study investigated the role of the morpho-semantic factor in the early processing of morphologically complex L2 words by changing the SOA. Our findings show even Chinese learners of English whose native language is typologically different from English can employ the rule-based decomposing mechanism. The decomposition is based on the interplay of morpho-orthographic and morpho-semantic information, adding new evidence to the assumption of parallel orthography-semantics activation in the early processing of L2 morphologically complex words.

Footnotes

Appendix

Appendix 3.

Semantically related condition.

Target	Related prime	Unrelated prime
Proof	Evidence	Recovery
Accuse	Charge	Period
Precious	Valuable	Ordinary
Dress	Skirt	North
Rich	Wealthy	Program
Endless	Infinite	Telegram
Field	Domain	Guilty
Faith	Belief	Option
Heaven	Paradise	Delivery
Outline	Summary	Younger
Urban	Village	Selfish
Selection	Choice	Battle
Doubt	Suspect	Miracle
Emphasize	Highlight	Represent
Verify	Confirm	Breathe
Equity	Justice	Current

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Open Fund of State Key Lab of Cognitive Neuroscience and Learning (CNLYB1309) to Baoguo Chen.

Author biographies

Qingge Zhang is a Master degree’s student from School of Psychology at Beijing Normal University. Her research focuses on phonological-orthographic processing during second language processing.

Lijuan Liang works as a lecturer at the School of English and Education at Guangdong University of Foreign Studies. Her research focuses on phonological-orthographic processing during second language processing.

Panpan Yao is a PhD student in Linguistic Department, Queen Mary University of London, working on comparative syntax, sentence processing and language acquisition.

Shanshan Hu is a Master degree’s student from School of Psychology at Beijing Normal University. Her research focuses on phonological-orthographic processing during second language processing.

Baoguo Chen is a Professor from the School of Psychology at Beijing Normal University. He mainly conducts research on L2 vocabulary learning, L2 semantic and syntactic processing and bilingual switching.

References

Baayen

R. H.

Davidson

D. J.

Bates

D. M.

(2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language, 59, 387–556.

Basnight-Brown

Chen

Shu

Kostić

Feldmann

(2007). Monolingual and bilingual recognition of regular and irregular English verbs: Sensitivity to form similarity varies with first language experience. Journal of Memory and Language, 57, 65–80.

Beyersmann

Castles

Coltheart

(2011). Early morphological decomposition during visual word recognition: Evidence from masked transposed-letter priming. Psychonomic Bulletin & Review, 18, 937–942.

Boudelaa

Marslen-Wilson

W. D.

(2005). Discontinuous morphology in time: Incremental masked priming in Arabic. Language and Cognitive Processes, 20, 207–260.

Brysbaert

New

(2009). Moving beyond Kučera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods, 41, 977–990.

Chu

S. C. R.

(1982). Chinese grammar and English grammar: A comparative study. Taipei, Taiwan: Commercial Press.

Clahsen

Neubauer

(2010). Morphology, frequency, and the processing of derived words in native and nonnative speakers. Lingua, 120, 2627–2637.

Diependaele

Duñabeitia

J. A.

Morris

Keuleers

(2011). Fast morphological effects in first and second language word recognition. Journal of Memory and Language, 64, 344–358.

Diependaele

Sandra

Grainger

(2005). Masked cross-modal morphological priming: Unravelling morpho-orthographic and morpho-semantic inﬂuences in early word recognition. Language and Cognitive Processes, 20, 75–114.

10.

Diependaele

Sandra

Grainger

(2009). Semantic transparency and masked morphological priming: The case of prefixed words. Memory & Cognition, 37, 895–908.

11.

Feldman

L. B.

O’Connor

P. A.

Del Prado Martín

F. M.

(2009). Early morphological processing is morphosemantic and not simply morpho-orthographic: A violation of form-then-meaning accounts of word recognition. Psychonomic Bulletin & Review, 16, 684–691.

12.

Feldman

L. B.

Soltano

E. G.

(1999). Morphological priming: The role of prime duration, semantic transparency, and affix position. Brain & Language, 68, 33–39.

13.

Feldman

L. B.

Soltano

E. G.

Pastizzo

M. J.

Francis

S. E.

(2004). What do graded effects of semantic transparency reveal about morphological processing? Brain & Language, 90, 17–30.

14.

Frost

Kugler

Deutsch

Forster

K. I.

(2005). Orthographic structure versus morphological structure: Principles of lexical organization in a given language. Journal of Experimental Psychology: Learning, Memory and Cognition, 31, 1293–1326.

15.

Giraudo

Grainger

(2001). Priming complex words: Evidence for supralexical representation of morphology. Psychonomic Bulletin and Review, 8, 127–131.

16.

Heyer

Clahsen

(2015). Late bilinguals see a scan in scanner AND in scandal: Dissecting formal overlap from morphological priming in the processing of derived words. Bilingualism: Language and Cognition, 18, 543–550.

17.

Kazanina

Dukova-Zheleva

Geber

Kharlamov

Tonciulescu

(2008). Decomposition into multiple morphemes during lexical access: A masked priming study of Russian nouns. Language and Cognitive Processes, 23, 800–823.

18.

Kirkici

Clahsen

(2012). Inﬂection and derivation in native and non-native language processing: Masked priming experiments on Turkish. Bilingualism: Language and Cognition, 1, 1–16.

19.

Liang

L. J.

Chen

B. G.

(2014). Processing morphologically complex words in second-language learners: The effect of proficiency. Acta Psychologica, 150, 69–79.

20.

Longtin

C. M.

Segui

Halle

P. A.

(2003). Morphological priming without morphological relationship. Language & Cognitive Processes, 18, 313–334.

21.

Lück

Hahne

Clahsen

(2006). Brain potentials to morphologically complex words during listening. Brain Research, 1077, 144–152.

22.

Marelli

Luzzatti

(2012). Frequency effects in the processing of Italian nominal compounds: Modulation of headedness and semantic transparency. Journal of Memory and Language, 66, 644–664.

23.

Marslen-Wilson

W. D.

(2007). Morphological processes in language comprehension. In Gaskel

(Ed.), The Oxford handbook of psycholinguistics (pp. 175–193). Oxford, UK: Oxford University Press.

24.

Marslen-Wilson

W. D.

Bozic

C. M.

Randall

(2008). Early decomposition in visual word recognition: Dissociating morphology, form, and meaning. Language and Cognitive Processes, 23, 394–421.

25.

Marslen-Wilson

W. D.

Tyler

L. K.

Waksler

Older

(1994). Morphology and meaning in the English mental lexicon. Psychological Review, 101, 3–33.

26.

McCutchen

Logan

Biangardi-Orpe

(2009). Making meaning: Children’s sensitivity to morphological information during word reading. Reading Research Quarterly, 44, 360–376.

27.

Morris

Frank

Grainger

Holcomb

P. J.

(2007). Semantic transparency and masked morphological priming: an ERP investigation. Psychophysiology, 44, 506–521.

28.

Neubauer

Clahsen

(2009). Decomposition of inflected words in a second language: An experimental study of German participles. Studies in Second Language Acquisition, 31, 403–435.

29.

Newman

A. J.

Ullman

M. T.

Pancheva

Waligura

D. L.

Neville

H. J.

(2007). An ERP study of regular and irregular English past tense inflection. NeuroImage, 34, 435–445.

30.

Pliatsikas

Marinis

(2013). Processing of regular and irregular past tense morphology in highly proficient second language learners of English: A self-paced reading study. Applied Psycholinguistics, 34, 943–970.

31.

Rastle

Davis

M. H.

(2003). Reading morphologically complex words: Some thoughts from masked priming. In Kinoshita

Lupker

S. J.

(Eds.), Masked priming: State of the art (pp. 279–305). New York, NY: Psychology Press.

32.

Rastle

Davis

M. H.

(2008). Morphological decomposition based on the analysis of orthography. Language and Cognitive Processes, 23, 942–971.

33.

Rastle

Davis

M. H.

New

(2004). The broth in my brother’s brothel: Morpho-orthographic segmentation in visual word recognition. Psychonomic Bulletin and Review, 11, 1090–1098.

34.

Raveh

Rueckl

J. G.

(2000). Equivalent effects of inflected and derived primes: Long-term morphological priming in fragment completion and lexical decision. Journal of Memory and Language, 42, 103–119.

35.

Rueckl

J. G.

Aicher

K. A.

(2008). Are CORNER and BROTHER morphologically complex? Not in the long term. Language and Cognitive Processes, 23, 972–1001.

36.

Silva

Clahsen

(2008). Morphologically complex words in L1 and L2 processing: Evidence from masked priming experiments in English. Bilingualism: Language and Cognition, 11, 245–260.

37.

Sonnenstuhl

Eisenbeiss

Clahsen

(1999). Morphological priming in the German mental lexicon. Cognition, 72, 203–236.

38.

Taft

(1981). Preﬁx stripping revisited. Journal of Verbal Learning and Verbal Behavior, 20, 289–297.

39.

Taft

Forster

K. I.

(1975). Lexical storage and retrieval of preﬁxed word s. Journal of Verbal Learning and Verbal Behavior, 14, 638–647.

40.

Ullman

M. T.

(2001). A neurocognitive perspective on language: The declarative / procedural model. Nature Reviews Neuroscience, 2, 717–726.

41.

Ullman

M. T.

(2004). Contributions of memory circuits to language: The declarative/ procedural model. Cognition, 92, 231–270.

42.

Ullman

M. T.

Pancheva

Love

Yee

Swinney

Hickok

(2005). Neural correlates of lexicon and grammar: Evidence from the production, reading, and judgment of inflection in aphasia. Brain and Language, 93, 185–238.