Frequency effects, the boringly powerful predictor of what your students will learn.

Nick Ellis spent twenty years arguing that the most boring variable in second language acquisition was also the most powerful. How often a learner encounters a form predicts how well they learn it. The usage-based revolution in linguistics took the field a while to catch up.

Frequency as the engine

Nick Ellis's 2002 paper (Frequency Effects in Language Processing: A Review with Implications for Theories of Implicit and Explicit Language Acquisition, Studies in Second Language Acquisition) was the field-defining statement. Ellis pulled together evidence across psycholinguistics, computational modeling, and SLA showing that nearly every level of language is sensitive to frequency. Words, collocations, constructions, syntactic patterns. The more often a unit appears in the input, the faster the learner processes it, the more accurately they produce it, and the more reliably they retain it.

This sounds obvious. The reason it was a controversy is that the dominant generative linguistics tradition had treated frequency as relatively unimportant. Chomsky's framework located grammar in innate Universal Grammar, with input serving mainly as a trigger for genetic endowment. Frequency, in that view, was peripheral.

The usage-based linguists (Bybee 2006, Goldberg 2006, Tomasello 2003) argued the opposite. Grammar emerges from accumulated experience with concrete utterances. Frequency is the central variable. Ellis applied the usage-based framework to L2 acquisition. The data largely sided with him.

What frequency actually means

Two kinds of frequency matter. They behave differently in learners.

Token frequency

How often a specific word or specific construction appears. The word tener has high token frequency in Spanish. So does the chunk tener que + infinitive. High token frequency produces fast, accurate processing. The learner gets good at the specific item.

Type frequency

How many different items follow the same pattern. The present-tense -ar ending has high type frequency: many different verbs follow it. High type frequency produces generalization. The learner extracts the pattern and applies it to new verbs.

Goldberg (2006, Constructions at Work, Oxford) summarizes the interaction: token frequency entrenches a specific form, type frequency abstracts a pattern. A learner with high token exposure to tener que + infinitive but no type variety will get good at that specific chunk and bad at productive uses of the modal pattern. A learner with high type variety extracts the modal pattern and can use it productively.

Frequency is a key factor at every level of linguistic analysis, and theories of language acquisition that ignore it are incomplete.

The Zipfian distribution

Word frequency in any natural language follows a Zipfian distribution. A small number of words account for most of the running text. The most frequent thousand words in Spanish cover roughly eighty percent of running text. The next two thousand cover another ten percent. The long tail (everything else) covers the remaining ten percent.

For acquisition, this means that the most frequent thousand words give the learner enormous receptive coverage relatively quickly. The next two thousand are critical for moving from intermediate to advanced. Past three thousand, each additional word adds less utility per word than the previous.

Paul Nation's vocabulary research (Learning Vocabulary in Another Language, Cambridge 2013) is the most practical statement of this. Nation argues for explicitly teaching the high-frequency word lists, letting the mid-frequency words come from extensive reading, and treating the low-frequency words as look-up vocabulary rather than memorization targets.

What this means for Spanish input planning

Three pedagogical implications.

The right thousand

Spanish has frequency lists. Davies's A Frequency Dictionary of Spanish (Routledge 2018) is the standard. The top thousand words include everything from ser and estar through household vocabulary, basic conjunctions, time expressions, and the common verbs. If your Spanish 1 curriculum is not roughly aligned with the top thousand by frequency, you are teaching low-utility words first.

Audit your unit one vocabulary. Count the entries that appear in the Davies top thousand. If fewer than seventy percent do, the unit was built for a textbook reason, not for utility. The fix is not to throw out the textbook but to weight your assessment toward the high-frequency vocabulary in each unit.

Frequency-weighted exposure

Time spent on each item should roughly track its frequency in real Spanish. A Spanish 1 student should be encountering ser, tener, and ir many hundreds of times before they finish the year. They should be encountering most lower-frequency food vocabulary far fewer times.

Many textbook units invert this. They give a vocabulary list with twenty equally-weighted items, treat them as equally important, and assess them with equal weight. The learner ends up over-trained on words they will rarely need and under-trained on words they will use every day.

Type-variety practice

For generalizable patterns (verb endings, adjective agreement, gender), expose students to many different types early. Twenty different -ar verbs in the present tense in week one is more useful than four different -ar verbs drilled hard. The type variety abstracts the pattern. The drilled token entrenches a chunk.

The frequency-formula relationship

High-frequency multi-word units (formulas) carry an outsized share of conversational fluency. Phrases like no sé, creo que, por supuesto, de acuerdo, vamos a ver, la verdad es que appear constantly in Spanish speech. Wray (2002, Formulaic Language and the Lexicon, Cambridge) and others have shown that L2 learners who acquire high-frequency formulas early sound much more fluent than learners with the same word counts who lack the formulas.

Textbooks rarely organize around formulas. They organize around grammar and topical vocabulary. A frequency-aware curriculum would teach high-frequency conversational formulas explicitly and early.

The downside of pure frequency

Frequency is not the only variable. Three caveats.

First, salience matters. Low-frequency forms that carry high communicative load (the subjunctive in some contexts, certain idiomatic expressions) can be acquired despite their lower input frequency because they are salient when they appear.

Second, frequency in input is not the same as frequency in instruction. A form that is rare in real Spanish (the future perfect, for example) but heavily drilled in instruction will be over-learned relative to its utility.

Third, learners need exposure to lower-frequency items for cultural and pragmatic reasons. Heritage vocabulary, regional dialect features, and specialized registers all matter even when their corpus frequency is low.

What this means for your Monday

Pick your next unit. Find a Spanish frequency list (Davies or any reputable online corpus tool). Mark each vocabulary item with its frequency rank. Sort the unit list by rank. The top ten items get more class time and more assessment weight. The bottom ten get less.

For at least three of the high-frequency items, look up the most common collocations they appear in. Teach the collocations alongside the word. Tener alone is useful. Tener que, tener ganas de, tener razón, tener hambre together are actual Spanish.

Audit your assessment. If the test weights all twenty vocabulary items equally, you are inadvertently training students to over-allocate study time to low-frequency items. Weight your test by frequency. The students will notice, and they will allocate their study time toward the high-utility items.