assimilated to the same native language sounds, however, are equally close to the native sounds or equally aberrant. In this situation, it is hard to differentiate them. This kind of discrimination is poor. The last four minor type “Both Unclassified” displays that the two non-native language speech sounds are felt as the speech-like sounds but cannot find any correspondence in the native language. The distinction can range from poor to very good, relying on their closeness to each other and to the native speech sounds.

2.1.2 SLM Model

Flege's speech learning model (SLM) (Flege, 1995) predicts that successful production of second language learning production  relies on the construction of new phonetic categories for the constituent in L2, but they don’t have any counterparts in the L1. There are seven hypotheses all together in SLM, among which three hypotheses are closely related to the effect of L2 speech production.

Firstly, sounds in L1 and L2 are related perceptually to one  another at a position-sensitive allophonic level, rather than at a more abstract phonemic level (Flege, 1995). It suggests that L2 language learners will connect an L2 phoneme to the nearest allophones rather than to a phoneme merely. The phonetic dissimilarities between L1 and L2 are either removed or assimilated to the first language phonological system not necessarily whereas maybe comprehended by bilingual. If L2 learners can be aware of the phonetic disparities between the L1 and L2 sound systems, they will be likely to establish a new category for the L2 sounds. Second, the sound hypothesis of the SLM suggests that “a new phonetic category can be established for an L2 sound that differs phonetically from the closest L1 sound if bilinguals discern at least some of the phonetic difference between the L1 and L2.” (Flege, 1995) Finally, a

sound will also be produced by L2 learners when the phonetic category of that phone is constructed.

Third, the sound hypothesis of the SLM shows that “the production of a sound eventually corresponds to the prosperity represented in its perceptual phonetic category representation.” (Flege, 1995)

2.2. Previous Research

2.2.1 English Consonants /n/ /l/ and /r/

A great number of previous researches have been conducted to account for the difficulty in acquiring English /n/ /l/ and /r/ for second language learners. For example, the study was carried out by Sheldon & Strange (1982), which demonstrated that English /r/ was identified correctly more than English /l/. Apart from them, Flege et al. (1995) tested that Japanese speakers were better in perception of English /r/ than English /l/. Similarly, Aoyama et al. (2004) investigated Japanese EFL learners’ perception of English /l/ and /r/ and found that English /r/ is perceptually less similar to Japanese /r/ than is English /l/. In Spanish, for example, the native Spanish speakers can find the corresponding counterparts for English consonants /l/ and /n/, which is easy for them to perceive these two consonants. (Garcia Lecumberri, M. L., & Cooke, M. 2006: 2445-2454).

2.2.2 The Relationship of Perception and Production

Some researchers hold that production is based on perception, and more accurate perception can contribute to more correct production. Ma Jia (2015) studied the patterns on perception and production of two sets phonemes /l/ - /r/ and /l/ - /n/ with English learns to lend support for the proposition of SLM that the production accuracy of L2 is based on the perception accuracy of L2. Borden, Gerber & Milsark (1983) detected the relationship between perception and production of English consonants  /l/and /r/ in Korean learners of English in an experiment that comprised production, identification, discrimination and a self-perception test. Compounding stimulus were used for the identification  and discrimination tests. One of the main results obtained was that self-perception develops earlier and may be a prerequisite for accurate production. And the research by Kluge et al. (2007) examined the perception and production of English /m/ and /n/ in code position by native speakers of Brazilian Portuguese which showed a positive correlation between the perception and production tests. Another study on the perception and production of Italian stop consonants by Austrian German learners carried out by Grasseger (1991) also supported the proposition that well-established perceptual categories do help accurate production. As Barry (1989) did for vowels, that perceptual tests might be a good tool to indicate production difficulties.