For many learners of a second language, multiple factors complicate learning the oral component of a new language. Firstly, they have to learn how to segment speech that they hear into phrases and words, so that they can decode what these mean. This is not a simple task, as speech is in fact a continuous acoustic signal, and the burden is on the hearers to know where to segment the signal. Furthermore, there is a great amount of variability – or lack of invariance – in second language speech. Coarticulation, where the context of the words can affect how the same sound is produced, can lead to variation within a single person’s speech. For instance, the /d/ is acoustically different in [di] and in [du] (and probably sounds quite different to the naked ear), as the vowel influences the sound of the preceding consonant. However, there are also variations in the nature of the sound signal received due to physiological reasons: gender, age, and even weight has been shown to affect the quality and frequency of speech.

Across speakers, regional accents can also make it difficult to understand what the speaker is saying. For example, second language French classrooms often teach based on the Parisian French accent, but second language learners can then face problems when trying to decipher the French of speakers from Marseille or Bordeaux, who have different accents. The context of understanding speech is also rarely perfect: there is often background noise which prevents hearer from obtaining the complete acoustic signals from speech. Hence, second language learners have to develop skills to normalize all this variation within and across speakers, whose speech they have to understand within imperfect and noisy settings.

Native speakers of a language develop particular faculties in order to manage the challenges of perceiving speech. They can use top-down processing, where contextual information helps them to guess what was said. For example, if you are in a noisy environment, and your friend tells you “I’m going to the toilet” while pointing in the direction of the toilet, even if you only hear the word “going”, you have sufficient information to guess, based on non-linguistic cues such as pointing and prior background knowledge of where the toilet is, that your friend is going to the toilet. Conversely, L2 learners have no such recourse, and it will be more difficult for them to understand the same sentence if placed in a similar situation.

Furthermore, categorical perception is developed by native speakers to aid in the discrimination of sounds as they are categorized by that particular language. This categorical perception could be different for second language learners whose native languages are dissimilar to the target language, such as Japanese. Native Japanese speakers learning English have been found to have problems differentiating between the English /r/ and /l/ (and also producing them), as Japanese does not possess an explicit distinction between the two sounds, which in fact exist on a continuum (Iverson et al., 2003).

As for speech production, there may be sounds that the learner has not produced before. Imagine if someone who has never spoken an African language now has to learn different kinds of clicks which are so prevalent in African languages like Xhosa (Lewis & Roux, 1996). This would certainly take some time and effort to acquire. Therefore, some sort of training is required to train the learner’s muscles that are used for speech, such that they can now produce these novel sounds, as researchers have tried to do with students learning English as a second language (Owolabi, 2012). Difficulty is increased when L2 learners acquire sounds in isolation, but then have to incorporate these sounds into the contexts of words, phrases, and sentences. Coarticulation implies that learners also have to be aware of when the sounds should be altered, but learning sounds in isolation does not provide this range. For example, in the phrase “report book”, native speakers tend to elide the /t/ sound, and second language learners who are unaware of that may pronounce it, which could make their speech sound odd to native speakers.

Lastly, perception and production do not have a direct relationship. Even if a second language learner is able to correctly discern L2 sound categories, accurate production of these categories is not guaranteed. After all, children have been found to be able to distinguish sounds not found in their native languages from native speech sounds after 8 months of age (Kuhl et al., 2014). Furthermore, native Cantonese speakers who are English second language learners in Hong Kong (Chan, 2014) produced English consonants that they perceived to be similar to their Cantonese counterparts poorly, even though they should have been easier to produce. They also had difficulties in producing certain sounds that they showed high accuracy in perceiving, like the consonant found in the English word “they” (Chan, 2014).