The interaction hypothesis, what conversation does that one-way input cannot.

Michael Long built on Krashen and went past him. Comprehensible input alone, Long argued, is not what drives acquisition. The work happens in the negotiation of meaning. Conversation, not lecture. The empirical record over thirty years has held up his argument.

The original claim

Michael Long's 1996 chapter (The Role of the Linguistic Environment in Second Language Acquisition, in Ritchie and Bhatia, Handbook of Second Language Acquisition) is the canonical statement. Long argued that interaction modifies input in ways that make the modified input particularly useful for acquisition. When a learner hits a comprehension breakdown in conversation, the partner reformulates, repeats, slows down, expands. The learner is now hearing a version of the message that has been tailored to their gap.

Krashen had argued that comprehensible input was sufficient. Long's update was that interaction produced more comprehensible input than non-interactive sources, and that the interactional modifications themselves did work the input alone could not. Negotiation of meaning was the active ingredient.

Negotiation of meaning, defined

Teresa Pica's 1994 paper (Research on Negotiation, Language Learning) gave the field its working definitions. Negotiation of meaning happens when a comprehension problem disrupts the flow of a conversation. The participants do something to repair it. Pica identified the moves.

Clarification requests

The listener signals that something was not understood. ¿Cómo? ¿Qué dijiste? ¿Puedes repetir? The speaker reformulates.

Confirmation checks

The listener checks their understanding by partially repeating with rising intonation. ¿Dijiste que fuiste al cine? The speaker confirms or corrects.

Comprehension checks

The speaker checks whether the listener has understood. ¿Entiendes? ¿Está claro? The listener signals understanding or requests clarification.

Reformulations and recasts

One participant offers a corrected version of the other's utterance. The recast research from the Lyster article shows where this fits in the broader feedback typology.

What the empirical record shows

Alison Mackey's 1999 paper (Input, Interaction, and Second Language Development, Studies in Second Language Acquisition) was the cleanest experimental test. Mackey compared four groups of L2 English learners: a group that received interactional input with negotiation, a group that received the same input without negotiation, a group that received modified input without interaction, and a no-interaction control. The negotiation group outperformed all other groups on subsequent production tests.

The Mackey and Goo 2007 meta-analysis pulled together every interaction study published through 2005 and confirmed the effect. Interaction with negotiation produces larger gains than non-interactive comprehensible input alone, with the effect strongest for grammatical accuracy and lexical acquisition.

It is proposed that environmental contributions to acquisition are mediated by selective attention and the learner's developing L2 processing capacity, and that these resources are brought together most usefully during negotiation for meaning.

Why interaction works when it works

Three reasons the interaction researchers have converged on.

Modified input is more comprehensible

The version of the message that emerges from a negotiation is tailored to the learner's actual gap. Generic comprehensible input cannot do this. Only interaction can produce input shaped to the specific listener at the specific moment.

Attention is focused

The breakdown forces the learner to attend. Generic input can wash over a distracted listener. A breakdown plus clarification request forces the listener back into focused processing.

Output is pushed

The learner has to say something. That brings in Swain's output hypothesis (covered in the output article). The push to produce reveals what the learner can do and what they cannot, and the negotiation around the gap supplies the input the learner needs.

What this means for classroom design

The interaction hypothesis is the strongest theoretical defense for pair and small-group work in a language classroom. Whole-class teacher talk is comprehensible input. Pair work is interaction with negotiation. The two do different cognitive jobs.

The teacher cannot negotiate one-on-one with every student for any meaningful amount of class time. The interaction hypothesis says students need that negotiation. The math works only if students do most of the negotiating with each other.

The Pica taxonomy of task types

Pica also identified which task types produce the most negotiation. Information-gap tasks produce the most. Decision-making tasks produce moderate amounts. Opinion-exchange tasks produce the least.

Information-gap tasks

Two students each have part of the information. To complete the task, they have to share. One has the map, the other has the destinations. One has the recipe, the other has the ingredients. One has the picture, the other has the description.

These produce dense negotiation because both students have to communicate clearly and confirm understanding to succeed. Both have a stake.

Decision-making tasks

Pairs or small groups have shared access to information and have to reach a decision. Plan the trip. Choose the restaurant. Solve the mystery. These produce some negotiation but less than info-gaps, because students can partially proceed without confirming each other's contributions.

Opinion-exchange tasks

Students share opinions on a topic without a shared goal. These produce the least negotiation because misunderstandings rarely block the activity. Students can keep talking past each other.

The friend-pair problem

Pair work works best when partners have a reason to negotiate. Best friends often skip negotiation. They already know each other's patterns. They guess through gaps. They switch into English for the parts they cannot do in Spanish. The activity produces less acquisition than intended.

Random pairing solves this. Same-strength pairing solves part of it. Mixed-strength pairing creates a different problem: the stronger student can carry the conversation and the weaker student can hide.

The pragmatic answer is to vary the pairing across the week, to design the task so both partners have to contribute, and to use information-gap or decision-making formats more than opinion-exchange formats when acquisition (rather than fluency practice alone) is the goal.

The heritage-non-heritage pairing question

Pairing a heritage student with a non-heritage student in a Spanish 3 class produces predictable problems. The heritage student carries the conversation. The non-heritage student processes a lot of input. The non-heritage student does not produce much. The heritage student does not learn anything new.

The interaction hypothesis suggests two fixes. One: pair students by ability for tasks where production matters most. The non-heritage student can produce in a same-level pair. Two: design tasks that draw on each student's strengths. Heritage students bring real-world knowledge, family expertise, vocabulary the L2 student does not have. The L2 student brings academic skills, reading strategies, sometimes more recent formal instruction. The task can be set up so both contribute genuinely.

What this means for your Monday

Audit the last week of pair work. Was the dominant task type information-gap, decision-making, or opinion-exchange? For most US Spanish classrooms, the answer is opinion-exchange (turn-and-talk about a topic). The opinion-exchange task is easy to set up and produces the least acquisition gain. The information-gap task is harder to set up and produces the most.

For next week, replace two opinion-exchange activities with two information-gap tasks. Track what happens to student talk time and to the language students reach for. The negotiation will increase. The acquisition gains follow the negotiation.