Developing Accurate and Fluent Speech

or why discussion clubs may not be helping your speech

Disclaimer: The bulk of this essay refers to two key models: Anderson’s theory of cognitive learning and Bialystok’s processing model. That being said, it would be best to take this post with a grain of salt and think of it as an exposition of the theory/model in practice.

On the first day of any course, I ask my students what their goals are. Responses vary, but the lowest common denominator is always improved speech. As useful as it is to have a goal, this one is unfortunately more ambiguous than most students realize. Is it more grammatically accurate speech they’re looking for, more native-like pronunciation, faster speech (etc.)? When prompted with these suggestions, students usually respond with a frustrated, “Yeah, all of that.”

With a mere 45 hours and half a textbook to get through, only so much time can be dedicated to speech, and practice is important. Thus, I suggest different ways students can get their oral communication fix outside of class. Despite the limited number of opportunities there are for real speaking outside of class, I normally only suggest discussion clubs when my employers tell me to promote it. Why’s this? The short version: to achieve accurate and fluent speech, a student needs accurate input, a clear understanding of how new language constructs are used, extensive practice, and appropriate feedback to avoid fossilization of inaccurate use…things I don’t see in abundance at discussion clubs. Let me explain.

Knowledge and Skill

We can break speech (oral language production) into two separate elements: knowledge of language (sometimes referred to as competence) and speaking as a skill (or control). This makes sense–knowing an arbitrary rule and using it properly are far from one and the same. Even though knowledge influences speech (in the beginning at least), we’re going to view them separately a la Bialystok’s processing model. This means we can rate any speech sample according to accuracy (how grammatically and contextually correct speech is) and fluency (how often the speaker pauses or hesitates while speaking). Following this model, we can rate speech as: inaccurate and hesitant, accurate and hesitant, inaccurate and fluent, and accurate and fluent*.

From these possibilities, students looking for improved speech are most likely looking at our fourth option–accurate and fluent speech. So how can we achieve that?

Accuracy and Declarative Knowledge

Most processing and learning models give us a pretty good first step to creating accurate speech: students need to know what correct/accurate speech is! There are two ways we can get a feel for this. Firstly, we could have a language rule explained to us, either from a teacher or a textbook (Bialystok refers to this as explicit knowledge), or we could rely on our listening skills/intuition and just pick up the language as we go (this is implicit knowledge; it relies on a boatload of input and is more common in immersion situations). After we’ve learned a language component, whether it be a grammar rule, construction, phrase or whatever, it can then be analyzed or not.

Analyzed knowledge doesn’t necessarily mean explicit knowledge. Analysis is when you make personal sense of a rule. Think of it like this: you were probably taught a full phrase in early-level language classes (explicit), but there’s a good chance you were given a basic translation and not explained how the phrase works. You can repeat this phrase and maybe even identify when someone uses it wrong, but you don’t know exactly why it’s wrong or why it’s right. This would be unanalyzed knowledge. On the flip side, you may have learned certain structures or phrases from hearing your parents or peers use them (implicit), but could only properly use them after thinking about what context you heard it in. By building that contextual definition, you’ve created analyzed knowledge. Bialystok puts a lot of emphasis on this. She basically says that without analyzed knowledge, a construct can only be repeated and not recombined to create new, spontaneous language. So, without analysis, we’re nothing more than parrots.

Regardless of whether language knowledge is acquired explicitly or implicitly, what’s important is that the learner can explain what is accurate and why/when it’s used. This is called declarative knowledge, and according to Anderson, it is essential to developing accurate (and spontaneous) speech. The idea is straightforward: a speaker needs to properly understand how to appropriately use a language construct before they can actually use it properly. If there is no declarative knowledge, then the speaker is just guessing and not reinforcing any kind of knowledge.

From Knowledge to Production

Now that we’ve covered declarative knowledge, we move on to stage two of Anderson’s model: production! This is what most people would call practice and it’s manifested in a number of ways: simple grammar drills, question prompts, etc. These are those exercise sheets and workbook activities you see every time a vocab or grammar topic is introduced. This is excellent for reinforcing knowledge, but it’s sterile and lacks practicality. Practice eventually needs to occur in spontaneous speech, but there’s a catch here: when it comes to using new elements outside of guided exercises, the speaker has to remember the construction, when to use it, and how to use it. This is something both Stevick and Spolsky identify as a major difficulty that language learners encounter–retention.

We can imagine our total language knowledge as a tool box with new items constantly being added. We’re not always quick to recognize opportunities to use new tools and, being the creatures of habit that we are, may instead fall back on more tried and true instruments. Thus, according to Faerch and Kasper, declarative knowledge requires activation in the speaker’s reception and production facilities before it can be used freely. So how do we activate this knowledge? With time and conscientious effort! This is a classic case of leading the horse to water. As much as students may not want to hear that they have to work, unless we’re dealing with run-of-the-mill language savants, learners need to keep declarative knowledge fresh (by studying…just an idea) and make an effort to use new components when the opportunity is present.

With enough production and effort, we will eventually move on to stage three of the model–what Anderson refers to as working memory and Bialystok calls automaticity.

Fluency and Fossils

Across SLA literature, you’ll find references to automatic and fluent language. For all intents and purposes, let’s say they’re roughly the same thing. When you become comfortable and familiar enough with a language component (we won’t get into long-term memory here, but keep that in mind), you’ll be able to use it naturally without much thought or hesitation. Essentially, by repeatedly using a construct, we create a habit of sorts and thus automatic speech. WE DID IT! Mission accomplished, right? Eeeeeeh, not exactly.

Remember those four potential speech outcomes? They were inaccurate and hesitant, inaccurate and fluent, accurate and hesitant, and accurate and fluent. We can figure inaccurate and hesitant speech occurs when a learner doesn’t fully understand the construct and is uncomfortable using it. This is natural when something is first learned. From here though, we have two possible paths for development: either the speaker will better understand the construction, the context it’s used in, and how it’s used (accurate and hesitant) or the speaker will become more comfortable using the construction, despite the fact that it’s used incorrectly (inaccurate and fluent). Although the former puts us on our way to achieving accurate and fluent speech, the latter leads us down the path to negative fossilization**.

Just like smoking or nail-biting, bad habits can be hard to break, and that’s what negative fossilization is. Although there’s much debate regarding fossilization, the general idea is that when we get so accustomed to speaking a certain way, it becomes near impossible to speak any differently (e.g. correctly). So what do we need to prevent this negative fossilization? You guessed it! Feedback.

This is a key element in Anderson’s cognitive learning model. As a new construct shifts from knowledge to usage, correction is required to ensure the learner properly understands and controls the new linguistic element. This can be done in a number of ways. Independently, speakers can monitor their own speech at the time of production and correct themselves, or they can reflect on their speech another time (as Spolsky explains, this could be checking a dictionary for pronunciation before or after a conversation). Feedback can also be given by a listener.

Vigil & Oller say feedback is when a listener provides input as a sign of approval/disapproval (affective) or understanding/a lack of understanding (cognitive). In practice, there are countless ways and degrees to which feedback can be given, but however it’s given, Ranta, Lyster, Schachter, Vigil, and Oller agree that some form of negative input is required in order to avoid fossilization. It makes sense that some form of feedback (or even explicit correction) is required to ensure language is accurate. Additionally, the learner should then make an effort to utilize this feedback to modify their own speech patterns and habits. What’s the point of getting feedback if it’s just going to be dismissed and forgotten five minutes later?

So…Discussion Clubs?

To quickly recap, we want accurate and fluent speech. To achieve this, we need declarative knowledge of a target construction. We’ll practice this construct, probably first in drills to make sure we properly understand it, and then make a conscious effort to use it in spontaneous speech. Monitoring our own speech, we’ll try to use the construct properly and correct ourselves if we use it incorrectly. We’ll also follow the feedback/input given to us by our speech partner and make corrections based on their responses. With enough time and practice, this construct will hopefully be used automatically in our speech. Okay. So then what’s my issue discussion clubs?

Actually, I like the idea of discussion club! For students who are restricted to a structured classroom environment, it may be the only chance they have for impromptu conversation. It may expose students to topics they may not come across in textbooks. It presents an exercise in creativity and a chance for learners to get their feet wet with new constructions, terms, etc. Since it’s still a controlled environment, it may be a better chance for students to gain confidence speaking a foreign language, which a study by Yule, Yanz, and Tsuda showed influences their accuracy. Plus, an environment where students can focus on form without feeling pressured is the perfect location for practicing their language and building fluency. The list goes on and on! And not just linguistically! A lot of discussion clubs are great social environments, too!

Unfortunately, this is a glass-half-full assessment. Let me preface this by saying I’m only familiar with the discussion clubs organized and run by nine different organizations in one city. Typically, there’s one teacher (usually a native speaker) and then seven to twelve language learners. We have ninety minutes. With nine people–eight students and one teacher–that’s 10 minutes of speech per person, give or take (especially since the teacher probably shouldn’t be as equal a participant and only a “host” to help lead the talk along). If there are twelve students, then we’re down to less than seven minutes a head. So, students have paid anywhere from nothing (yay!) to $10 USD (boo) for around ten minutes of actual speech…plus the time to get to the location and back and the event itself.

For students who go because they just want to listen, then we’re looking at more speech time for the other students, but that’s not necessarily a good thing for the listener. The discussion clubs I have experience with don’t put any emphasis on feedback. Either the hosts don’t want to interrupt the flow, don’t know how to provide constructive feedback, or don’t care. Additionally, at least 75% of the people who attend (the ones who just go to “maintain their language”) often dismiss feedback or make no effort to correct their speech, thus reinforcing their fossilized bad habits. Granted, if the learner’s goal is communication and doesn’t care about accuracy or form, then sure, why pay any attention to what the host says. Those who are just listening though are getting a front row seat to unchecked language. This means the incorrect language could, as per White, affect their speech. In this case, listeners would be better off staying at home and listening to a roundtable discussion with a transcript. Or a podcast. Or a TV show (and you all know how I feel about that).

Another issue that I’ve come across is often with the hosts. If their goal is just to keep a conversation alive for 90 minutes, then great. Issues arise when language that isn’t necessarily inaccurate is corrected. Speaking about English, due to the number of varieties that exist, natives may consider unfamiliar varieties incorrect since they’re spoken by a non-native (this is another Spolsky argument). This is something I am guilty of from my early days teaching (I corrected a girl who used the word fringe even though that is an A+ British word for what we call bangs).


Achieving accurate and fluent speech is neither quick nor easy. It requires some study, some work, and lots of practice. More importantly, it requires proper practice. Along the same lines that doing physical exercise improperly could hinder your development or even cause personal harm, improper language practice could damage your accuracy. Although opportunities for practice should be sought, they must provide for monitoring, feedback, and reflection. Even though discussion clubs provide a great opportunity for this kind of practice, due to the limited amount of time students can spend talking, attending once a week without supplement will doubtfully significantly improve fluency, and due to the speech of other attendees and depending on the host, meetings could cause more bad than good. If you’re looking for practice, there are other alternatives out there, or simply find a good discussion club. Make sure the host corrects speech, monitor your own speech, and be attentive. There is a lot that can be gained in an hour and a half, provided the right effort is made.

* Inaccurate and hesitant wasn’t originally listed in the literature, but I find it appropriate to include.

** My use of fossilization follows the definition set forth by Vigil & Oller.


  1. Anderson, J.R. 1980. Cognitive Psychology and Its Implications. San Francisco: Freeman. [170]
  2. Bialystok, E. 1978. ‘A theoretical model of second language learning.’ Language Learning 28: 69–84. [2, 47, 62]
  3. Bialystok, E. 1981. ‘The role of linguistic knowledge in second language use.’ Studies in Second Language Acquisition 4: 31–45. [41, 48]
  4. Ranta, L. and R. Lyster. 2007. ‘The Awareness-Practice-Feedback sequence’ Practice in a Second Language: Perspectives from Applied Linguistics and Cognitive Psychology. DeKeyser, R. New York: Cambridge. 141-160.
  5. Faerch, C. and G. Kasper. 1986. ‘Cognitive dimensions of language transfer’ in E. Kellerman and M. Sharwood Smith (eds.) 1986: 49–65. [49]
  6. Schachter, J. 1983. ‘Nutritional needs of language learners’ in M.A. Clark and J. Handscombe (eds.): On TESOL ‘82: Pacific Perspectives on Language Learning and Teaching. Washington, DC: TESOL. [188, 190]
  7. Spolksy, B. 1989. ‘Conditions for Second Language Learning. Introduction to a general theory.’ Oxford: Oxford University Press.
  8. Stevick, E.W. 1986. Images and Options in the Language Classroom. Cambridge: Cambridge University Press. [169, 199–200]
  9. Vigil, N. and J.W. Oller Jr. 1976. ‘Rule fossilization: a tentative model.’ Language Learning 26: 281–95. [189]
  10. White, L. 1987. ‘Against comprehensible input: the input hypothesis and the development of second-language competence.’ Applied Linguistics 8: 95–110. [192]
  11. Yule, G., J.L. Yanz, and A. Tsuda. 1985. ‘Investigating aspects of the language learner’s confidence: an application of the theory of signal detection.’ Language Learning 35: 473–88. [116(n 8), 169]

Leave a Reply

Your email address will not be published. Required fields are marked *