Why do learners – in the same essay – sometimes make an error in the use of a specific target-language structure and sometimes they don’t?

Teenage girl (16-17) lying on bed, writing, close-up

This morning, whilst correcting Spanish essays written my year 8 students (12-13 year olds if you are not familiar with the British school system) one mistake attracted my attention : a girl had written – on the same line, but within different sentences – ‘llevé una camiseta’ (I wore a T-shirt) and ‘llevo una camiseta’ (I wear a T-shirt) to mean, in both cases, ‘I wear a T-shirt’.  When asked to self-correct, she noticed her mistake immediately and changed the ‘é’ in ‘llevé’ to ‘o’ (a sign that she had declarative knowledge of the first person of the Preterite and Present in Spanish).

But why would a student produce the present tense correctly in one sentence and not in another within the same essay? And how could my year 8 student get the verb right on the second instance when she had just got it wrong a few words before on the same line?

As I often do in my one-to-one corrective conferences, I asked the student in question why she thought she had made that mistake. She shrugged and said: “No idea, sir’. Maybe I was tired”. A plausible explanation considering that she had written a long essay and that the error occurred in last paragraph. But how can tiredness cause such mistakes?

The answer to this question relates to what applied linguists call Interlanguage Variability, a widely documented phenomenon that causes frustration to a lot of teachers but which is actually a developmental feature of L2 acquisition. What is Interlanguage variability? What causes it?

To fully understand this phenomenon, the reader will benefit from getting acquainted with two important concepts: Interlanguage and Spread of Activation. I will take for granted that the reader is familiar with the concept of Working Memory, already explained in some detail in a previous post (see below: “Why do our learners get prepositions, articles and verb and adjectival agreement wrong?”). Please note that in what follows I will focus only on the cause of variability which, in my view, are more relevant to teachers operating in explicit foreign language instruction settings and that I will not venture into sociolinguistic theories of the likes of Labov’s nor into nativist accounts of the phenomenon.


Interlanguage is the name given by Selinker (1972) to the internal representation the L2-learner builds of the target language system in his/her Long Term Memory. How does s/he build it? Mainly through hypothesis-testing, often using his/her dominant language as a reference framework in an attempt to decode and make sense of the target foreign language. Since L2-acquisition occurs through trial and error, Interlanguage is not an exact system, but rather an approximation of the target language system one is acquiring.

It should be noted that the cognitive and affective feedback the learner receives from the target language speakers/knowers plays a pivotal role in the construction of the Interlanguage system, as it will ultimately determine which Interlanguage forms will be automatized and acquired. So, if a given Interlanguage form receives a lot of positive cognitive and affective feedback from the environment, it will eventually be internalized after the brain will have repeatedly been given reassurance that it is accurate.

What often happens, though, during the early stages of L2 acquisition, is that learners do not always receive consistently negative/positive cognitive or affective feedback on their errors; and even when they do receive it, it doesn’t necessarily follow that they will internalize it. This happens for a number of reasons to do with the corrective approach used in the classroom (e.g. selective or no correction); its quality (e.g. ambiguous feedback); strong interference from their first language which makes the Interlanguage structure more resistant to correction; etc.

Moreover, when students engage in unmonitored L2-production (in or outside the classroom), as happens in the course of unstructured communicative activities, their output is likely to contain more errors.

Although errors made at this stage are not automatized immediately, they will not be discarded by the brain straight away either, especially when they are repeated several times – and errors due to L1 transfer are likely to occur quite frequently at the early stages of L2 learning.  Hence, it is very common for the Interlanguage of an L2 learner to ‘contain’ more than one representation of a given target language structure: the correct one and one or more incorrect ones. Example: ‘I went’ in French is ‘je suis allé’, however, L2 students often say ‘j’ai allé’ at the early stages of French acquisition because they overgeneralize the dominant way of forming the Perfect Indicative in French. These two forms ‘j’ai allé’ and ‘je suis allé’ often coexist in L2 learners’ of French Interlanguage and compete with one another for retrieval. I will come back to this example. Now with this in mind let us look at the concept of ‘spread of activation’.

Spread of activation and Variability from processing inefficiency

When we are attending to a task, like forming the Perfect tense of ‘Aller’ in French, as in the above example, Working Memory will have to retrieve from Long-term Memory the correct match for ‘I went’ in French. As Working Memory attends to this tasks, every single bit of information (lexis, grammar, imagery)  related to the concept ‘I went’ stored in our Long Term Memory gets activated. ‘Electrical impulses’ run through semantic memory’s neural networks and the information or ‘nodes’ along the network get more or less activation based on the strength of their associations with the proposition we mean to ‘translate’ into French – the so-called ‘fan effect’. The items along the activated neural networks which will receive the greatest activation will be ‘‘j’ai allé’ and ‘je suis allé’ and, possibly, in my experience, ‘j’allé’. Which one of the three forms will be retrieved and used in the written/oral performance will depend on the ‘weight’ of each form (i.e. the strength of the memory trace) and on the context.

If the learner knows the correct French translation of ‘I went’ and Working Memory is not experiencing cognitive overload thereby having enough free space to monitor the output, even though s/he might have an initial moment of indecision due to the concurrent activation of the other two activated forms, s/he will be likely to apply the correct Interlanguage form. However, if his/her Working Memory is experiencing cognitive overload (processing inefficiency) due to a challenging task-in-hand, in the absence of close monitoring, any of the three forms may be retrieved (pretty much randomly) if their ‘weights’ are more or less equivalent. Hence the importance, at the early stages of learning, not to engage in overly unstructured oral or writing tasks.

Variability as caused by formulaic language

Variability can also be caused by formulaic-language learning that is to say the acquisition of unanalyzed chunks or set phrases memorized without really knowing what each constituent of the phrase actually means or how the grammar rules which ‘holds’ them together actually work. Thus, if a learner uses ‘je suis allé au cinema’ correctly in a written piece because s/he has learnt that sentence as an unanalyzed chunk, it will not mean that s/he masters the use of the Perfect Tense of verbs requiring the auxiliary ‘Etre’ in the Perfect Tense. Hence, when, a few lines below, in the same essay, s/he translates ‘I went’ incorrectly in a different context (e.g. I went to the park) we should not be particularly surprised by the occurrence of variability.

Variability as caused by learner strategies

Variability can also be caused by the learner’s attempt at testing a specific hypothesis they formulated about a given target language structure. Let us look at Muskaan’s hypothesis-testing strategy. Muskaan is a year 9 student of Spanish I teach who, today, told me that when she is not sure whether her assumptions about how to use a given structure are correct, tries them all out deliberately in order to get feedback from me as to which one is correct. In the essay we were marking together today, for instance, she had used a conditional and an imperfect form to translate two very similar sentences which should have required the imperfect. She wanted to tested the hypothesis that, just like in English you would use the conditional tense in sentences like “when I was young I would play the guitar in my free time’ one can do the same in Spanish. In Muskaan’s case, the retrieval of the two concurrent Interlanguage forms is not automatic / subconscious, but is triggered by a deliberate risk-taking strategy.

Risk-taking is another frequent cause of variability in our learners’ output and a phenomenon that must not be discouraged as it has great potential for learning.

In conclusion, Variability is a complete normal phenomenon that should not cause us too much frustration, even when it seem to be caused by our teaching. The most important implication of this phenomenon for the MFL classroom is that we need to be cognitively empathetic with our learners when we find this kind of mistakes and while addressing them through appropriate remedial learning, we must not stigmatize them. Secondly, teachers must give students enough time to monitor their output and encourage them to edit their written work carefully and in ways which lessen the cognitive load on their Working Memories (as the problem which triggered the error in production is likely to hinder its detection whilst proofreading). One such strategy is to have several runs through the same text, each one aimed at checking a particular type of item at a time (e.g. first time, adjectival agreement; second time, verb agreement; third time, omissions of copulas; fourth time ‘small function words’). Sentences that are particularly long and require complex processing should be dealt with by investing more time and focus.

Why do our learners often get prepositions, articles and verb endings wrong?


This article was prompted by a question (the one in the title) a colleague asked me recently at finding lots of mistakes in their students’ essays relating not solely to prepositions, but also to definite/indefinite articles, copulas (e.g. is and are) and other function words. The answer to that question is relatively simple, but in order for the reader to fully grasp its implications for classroom instruction, one has to first get acquainted with the concept of Working Memory and Executive Control.

Working Memory

To put it as simply as possible – since one of my colleagues keeps complaining about the complex jargon in my blogposts – Working Memory is the space activated in our brain when we process information, what in the old days was called Short-Term memory. Working Memory is a ‘buffer’ between the outside world and Long-Term memory which ‘holds’ any information we are trying to decode and retrieves the information we need from Long-term memory which we need to carry out the task-in-hand (e.g. writing a sentence or understanding a text). So, for example, when we are writing a sentence in a foreign language, Working Memory is the ‘place’ along our neural network in which we actually construct that sentence (i.e. where we choose the words we need from our mental lexicon, arrange them together in a grammatically correct sequence, make sure the spelling is correct and edit the final product).

Working memory has very limited channel capacity, in other words can only store a limited number of images, words and numbers at any one time and unless we keep rehearsing it, the information will be lost easily after a few seconds only (a word stays in our brain only 2-3 seconds unless we make a conscious effort to retain in through rehearsal). That is why, in order to keep a phone number in our head as we frantically try to key it in our phone we need to repeat it in our heads (or rehearse it) a few times. Miller’s (1965) magic number 7+/- 2 indicates the number of digits we can hold in our Working Memory at any one time – a very short number indeed.

The challenges posed by foreign language writing

Writing in a foreign language is much harder than a lot of us may think, especially under communicative pressure. Let us have a closer look at how a sentence is produced in writing. First of all, it is important to point out that the starting point, both in the first and the second language is a Proposition, in other words a representation in our brain (in Semantic Memory to be precise) of the concept or idea we are trying to convey. A Proposition, unlike what we may intuitively think, is not made up of words, thus, the brain has to translate into words, whether we are operating in the first language or second language.

According to Cognitive research (e.g. Cooper and Matsuhashi, 1983), the Translation process consists of four stages: Wording, Presenting, Storing and Transcribing. In the first stage, the brain transforms the Propositions into words (lexis). Although at this stage the pre-lexical decisions the writer made at earlier stages and the preceding discourse limit lexical choice, Wording the proposition is still a complex task: ‘the choice seems infinite, especially when we begin considering all the possibilities for modifying or qualifying the main verb and affected nouns’ (Cooper and Matsuhashi, 1983: 32).

Once s/he has selected the lexical items needed, the writer has to tackle the task of Presenting the proposition in standard written language. This involves making a series of decisions in the areas of genre and grammar. In the area of grammar, Agreement and Tense will be the main issues, especially in languages like French, or German where a lot of permutations are required.

The proposition, as planned so far, is then temporarily stored in Working Memory while Transcribing takes place. Propositions longer than just a few words will have to be rehearsed and re-rehearsed in Working Memory for parts of it not to be lost before the transcription is complete.

The limitations of Working Memory create serious disadvantages for unskilled writers. Until they gain some confidence and fluency with spelling, their Working Memory may have to be loaded up with letter sequences of single words or with only 2 or 3 words (Hotopf, 1980). This not only slows down the writing process, but it also means that all other planning must be suspended during the transcriptions of short letter or word sequences.

The physical act of Transcribing the fully formed proposition begins once the graphic image of the output (what the sentence physically looks like) has been stored in Working Memory.

In L1-writing the decisions taken at any of the four stages outlined above are taken automatically, thereby occupying little or no space at all in Working Memory. However, in L2-writing, especially in beginner to intermediate writer, every decision will take a lot of Working Memory space, making the process slow, cumbersome and difficult to monitor because the process happens mostly consciously.

Hence, the adaptive response of the brain, especially in beginner writers, is to prioritize the most important features of each proposition (the principle of ‘Saliency first’ being at play here), i.e. : the items that are most important in terms of conveying the intended meaning. The most semantically salient elements will include mainly: Nouns, Verbs and Adjectives. Function words, which carry considerably less meaning, will be relatively neglected by Working Memory’s attentional systems as, let’s face it, even if the writer gets them wrong, they won’t impede comprehension massively (example: whether I say, in French “Je vais au cinema’ or ‘je vais à la cinema’ I will be readily understood by a reader/listener).

This phenomenon is exacerbated by linguistic distance between the first language and the target foreign language. For instance gender (masculine and feminine) as well as verb endings are not likely to be perceived as salient by an English native speaker (as they do not exist in their language), which means that they are likely to be less monitored.

The less proficient the foreign language writer is and the less time he has to monitor his/her output, the more likely he/she will be to make mistakes with function words. Hence, errors are bound to be even more frequent in oral performance, where the self-monitoring capacity of Working Memory is drastically reduced compared to the written medium.

As the learner becomes more proficient, his/her ability to juggle the demands posed to his/her Working Memory by the processes outlined above will increase. This is due to the fact that with a lot of writing practice in the target foreign language a lot of sub-processes become automatized and require only peripheral attention, freeing up Working Memory space. This enhanced processing efficiency will also allow for more accuracy, too, in the production of less salient features unless Error Fossilization throws the spanner in the works.

The danger of fossilization

When errors go unmonitored a bit too often, they become automatized and it is very difficult to ‘unlearn’ or eradicate. Mukkatesh (1986) found that despite many remedial interventions such errors cannot be eliminate at all from L2 learners’ Interlanguage. This phenomenon, called by Selinker (1972) Fossilization, is obvious in a lot of foreign language speakers, especially when it comes to pronunciation; that is why, according to Selinker, only 5 % of foreign language speakers can be said to sound 100% native-like. Their second language will always contain some fossilized item. A very good friend of mine, for instance, speaks perfect English, with accurate pronunciation and grammar and an impressive lexical repertoire wider to that of an average native speaker; however, he cannot help voicing the ‘p’ in the word ‘psychology’ (influence of his first language: Italian), despite many corrections. Such is the power of Fossilization.

Function words and any other less salient L2 features (e.g. gender, plural and verb endings and minor pronunciation inaccuracies) are particularly amenable to fossilization as they are more likely to go unmonitored and uncorrected. Therefore, the danger is that when learners do not get enough negative (cognitive) feedback at the early stages of L2 acquisition, they are likely to fossilize mistakes with the above L2 structures and to keep making these mistakes all the way to A-Level and university – as I have often witnessed in my university lecturer days.

Communicative language teaching, especially in its strong version, by prioritizing fluency over accuracy, often leads to fossilization (and pidginization) especially when the students are asked to perform in unstructured oral practice, at a level of proficiency they are not developmentally ready for and under communicative pressure. (Skehan, 1994)

Implications for MFL teaching


The implications for the MFL classroom are manifold but hinge mainly on the teacher’s pedagogy and on the course end-goals. If we are teaching GCSE level students and we are happy for them to make a few minor mistakes as far as they can convey their intended meaning effectively, we should not worry too much about error and we can exercise a relatively high degree of tolerance. However, if we are dealing with individuals who want to make language their career and become one day interpreters, translators or teachers, then the attitude has to be less lax and mistakes with articles, prepositions, copulas and gender agreement WILL matter.

If we do want to address this issue radically, we need to keep students focused on the importance of accuracy from the very early stages of language acquisition whilst keeping the main focus of our teaching on the development of fluency. This is not easy, even for experienced teachers. Editing instruction – through games, quiz and other fun activities – should become part of almost every lesson (through snappy starters or plenaries, for example) to remind students of the importance of accuracy and to raise their awareness of which mistakes are more likely to occur at their current level of proficiency.

More importantly, the written tasks we involve are students in must pitched to the correct level, especially in terms of the cognitive challenges they pose to an inexperienced writer. If we do not, we are likely to engender more error than we and the students can effectively deal with in the remedial phase. Fluency, as I said above, has priority, it is true; however, fluent output that is rife with errors can be stigmatizing and irritating for the reader/interlocutor and we need to be aware of that in a global era in which, more than ever before, our learners are more likely to use the target language in the workplace.

Finally, Error correction – or rather Error remediation – can also play an important role if it engages the learners in a sustained long-term self-monitoring process initially moderated by the teacher which aims at focusing them on their most frequent mistakes.