Frequency Hypothesis

The Frequency Hypothesis proposes that the frequency of linguistic forms in the input is the primary determinant of acquisition order and ease of learning. The more often a learner encounters a form, the earlier and more robustly it is acquired. Nick Ellis (2002) provided the most comprehensive articulation of frequency-based accounts of SLA, demonstrating that language processing is intimately tuned to input frequency across every level of linguistic structure.

Core Claims

1. Input frequency drives acquisition — forms that occur more frequently in the input are acquired earlier
2. Frequency effects operate at every level — phonology, phonotactics, lexis, morphosyntax, formulaic sequences, and syntax all show frequency effects
3. Language learning is fundamentally statistical — learners are sensitive to the distributional properties of the input and extract patterns through probabilistic learning
4. No innate grammar module is required — frequency and distributional learning, combined with general cognitive mechanisms, can account for acquisition without positing Universal Grammar

Evidence

Ellis (2002) reviewed frequency effects across multiple domains:

• Phonotactics — learners acquire phonotactic patterns (permissible sound sequences) in proportion to their frequency in the input
• Lexis — high-frequency words are recognised faster and acquired earlier
• Morphosyntax — the acquisition order of grammatical morphemes correlates with their frequency in the input (though not perfectly — see criticisms below)
• Formulaic language — high-frequency chunks are stored and retrieved as wholes rather than being generated from rules
• Grammaticality judgements — learners are more accurate at judging the grammaticality of sentences containing frequent constructions

Relationship to Other Theories

The Frequency Hypothesis is the empirical backbone of Usage-Based Theory and Emergentism. If linguistic knowledge emerges from input exposure rather than from innate specification, then the statistical properties of the input — above all, frequency — become the primary explanatory variable.

It also connects to Connectionism, which models language learning as the strengthening of neural connections through repeated exposure to patterns. Connectionist models demonstrate that frequency-sensitive learning mechanisms can acquire complex linguistic behaviour without explicit rules.

The hypothesis challenges the Natural Order Hypothesis by offering a different explanation for acquisition orders: structures are acquired in a particular order not because of an innate syllabus, but because of their frequency and salience in the input.

Criticisms

• Frequency alone is insufficient — some high-frequency forms (e.g., English third-person -s) are acquired late, while some low-frequency forms are acquired early. Salience, communicative value, and L1 influence also matter.
• Type vs. token frequency — type frequency (how many different lexical items appear in a construction) and token frequency (how often a specific item appears) have different effects. Ellis (2002) acknowledged this distinction.
• Interaction with other factors — frequency interacts with perceptual salience, functional importance, and L1 transfer. A purely frequency-based account oversimplifies.

Teaching Implications

• High-frequency vocabulary and constructions should be prioritised in syllabus design
• Input Enhancement can compensate for low perceptual salience of frequent but non-salient forms (e.g., articles, morphological endings)
• Extensive reading and listening — which maximise input exposure — are supported by frequency-based accounts
• Recycling and spaced repetition align with the logic of frequency effects