From Exposure to Expectation: Frequency, Surprisal, and Language Across Development in Spanish

Francisco Portillo López

cs.CLAug 23, 2026

From Exposure to Expectation: Frequency, Surprisal, and Language Across Development in Spanish

Authors: Francisco Portillo López

Organizations: Universidad de Navarra, Pamplona, Spain

Abstract

Surprisal, the negative log-probability a language model assigns to a word given its preceding context, reliably predicts adult reading times. Does it contribute as much to explaining when children acquire individual words? Frequency reflects a learner's cumulative exposure to a word, whereas surprisal reflects how predictable a single occurrence is given its context. We investigate this question across two corpus-based studies of Spanish. In Study 1, we modeled age of acquisition (AoA) for 225 Spanish nouns using lexical frequency and contextual diversity from child-directed speech, plus surprisal from three language models differing in architecture and training language (BETO, BERTIN, mGPT). Frequency strongly predicted AoA (r=-.597, p<.001); surprisal added little beyond frequency and word length, including in a naturalistic-context analysis. In Study 2, we modeled adult fixation durations in the Chilean Spanish subsample of the Multilingual Eye-movement Corpus (MECO Wave 2), using mGPT surprisal alongside two independent frequency measures. Surprisal robustly predicted longer fixation durations after controlling for frequency and word length, consistent across both frequency sources. A matched word-type-level comparison showed the surprisal-behavior association was stronger in reading than in acquisition (z=3.63, p<.001). The findings suggest cumulative lexical exposure and contextual predictability play different roles across the language trajectory: frequency is particularly informative about when early lexical representations are acquired, whereas surprisal captures moment-to-moment processing difficulty in an already-established linguistic system. We discuss this pattern in relation to usage-based and entrenchment-based accounts of lexical development and to the evaluation of language models as models of human language behavior.

Figures & tables

Predictor	$\rho$	$p$	Context
Frequency (CDS)	$-.597$	$<.001$	Corpus
Contextual diversity	$-.564$	$<.001$	Corpus
Contextual diversity, partialled for frequency	$.028$	$.680$	Corpus
BETO surprisal	$-.036$	$.590$	Standardized frames
BERTIN surprisal	$.102$	$.128$	Standardized frames
mGPT surprisal	$-.048$	$.469$	Standardized frames

Table 1: Spearman correlations between distributional predictors and age of acquisition in Study 1. Sample sizes vary slightly across analyses because of missing corpus or surprisal estimates. The frequency and contextual-diversity analyses use

N=224

, whereas standardized-context LLM analyses use

N=225

and the natural-context mGPT analysis uses

N=224

Figure 1: Frequency in child-directed speech predicts age of acquisition for 225 Spanish nouns. The association was negative and significant (Spearman $r=-.597$ , $p<.001$ ). Selected early- and late-acquired nouns are labeled for illustration.

Figure 2: Corpus-based measures and LLM-derived surprisal as predictors of age of acquisition in Study 1. Dashed lines represent linear trends.

Model	$R^{2}$	$\Delta R^{2}$	$\beta_{\text{surprisal}}$	$p$
Frequency + word length	.379	–	–	–
+ BETO surprisal	.381	.0023	.048	.368
+ BERTIN surprisal	.385	.0067	.082	.124
+ mGPT surprisal	.380	.0006	$-.024$	.653

Table 2: Incremental variance in age of acquisition explained by standardized-context LLM surprisal beyond CDS frequency and word length in Study 1.

Sample	$N$	Baseline $R^{2}$	Full $R^{2}$	$\Delta R^{2}$	$\beta_{\text{mGPT}}$	$p$
All valid natural contexts	224	.3677	.3685	.0008	$-.0209$	.595
$\geq 5$ contexts per noun	154	.3135	.3211	.0076	$-.0737$	.197

Table 3: Incremental predictive value of natural-context mGPT surprisal for age of acquisition. Models include CDS frequency and word length as baseline predictors.

Figure 3: Standardized fixed effects predicting log total fixation duration in the MECO Wave 2 Chilean Spanish sample. Error bars show 95% confidence intervals.

Predictor	Model A (EsPal)	Model B (SUBTLEX-ESP)
Surprisal (mGPT)	$0.0114^{***}$ (SE $=.00067$ )	$0.0114^{***}$ (SE $=.00066$ )
Lexical frequency	$-0.0226^{**}$ (SE $=.0070$ )	$-0.0294^{***}$ (SE $=.0068$ )
Word length	$0.0581^{***}$ (SE $=.0025$ )	$0.0540^{***}$ (SE $=.0028$ )
Intercept	$5.349^{***}$ (SE $=.055$ )	$5.359^{***}$ (SE $=.048$ )
Item variance (word type)	$0.0200$ (SD $=.142$ )	$0.0198$ (SD $=.141$ )
Participant variance	$0.0354$ (SD $=.188$ )	$0.0354$ (SD $=.188$ )

Table 4: Mixed-effects model coefficients predicting log total fixation duration in MECO Wave 2 Chilean Spanish, with crossed random intercepts for participant and word type (item).

Figure 4: Graphical summary of the main findings from Studies 1 and 2. Study 1 shows that lexical frequency in child-directed speech strongly predicts age of acquisition (AoA) for Spanish nouns, whereas LLM-derived surprisal provides little additional predictive value beyond frequency and word length. Study 2 shows that mGPT surprisal predicts longer adult fixation durations during reading after controlling for lexical frequency and word length. Together, the findings illustrate different predictive roles for cumulative lexical exposure and contextual predictability across the Spanish language trajectory.

Surprisal Theory

See all 16 papers

May 7, 2026

The Frequency Confound in Language-Model Surprisal and Metaphor Novelty

Language-model (LM) surprisal is widely used as a proxy for contextual predictability and has been reported to correlate with metaphor novelty judgments. However, surprisal is tightly intertwined with lexical frequency. We explore this interaction on metaphor novelty ratings using two different word frequency measures. We analyse surprisal estimates from eight Pythia model sizes and 154 training checkpoints. Across settings, word frequency is a stronger predictor of metaphor novelty than surprisal. Across training stages, the surprisal--novelty association peaks at an early stage and then falls again, mirroring a similarly timed increase in the surprisal--frequency association. These results suggest that the often-reported optimal LM surprisal settings may incorrectly associate contextual predictability with metaphor novelty and processing difficulty, whereas lexical frequency may be the major underlying factor.

Omar Momen, Sina Zarrieß

CRC 1646 – Linguistic Creativity in Communication · Bielefeld University, Germany

Language Modeling Computational Psycholinguistics Surprisal Theory

Ask PDF

Oct 4, 2026

A Systematic Analysis of the Predictive Power of LM Surprisal in Reading Chinese

This study analyzes the predictive power of LM-derived, token-level surprisal on Mandarin Chinese reading times. We first propose the Shortest Matching Sequence (SMS), an alignment scheme that maps between the word segmentation assumed by eye-tracking corpora and the LMs' subword tokenization, as the two tokenizations often disagree in the context of Mandarin Chinese. Then, using a suite of Chinese-Pythia models (14M-1.4B) trained on scratch with 30B tokens, we examine how well surprisal predicts first fixation duration, gaze duration, and total reading time in three paragraph-level eye-tracking corpora of Mandarin Chinese (GECO-CN, HKP, and MECO). Contrary to previous null findings, our results show that surprisal is predictive of Chinese reading times. However, whether predictive power scales with model size and the amount of training is corpus-specific: bigger models predict better in GECO-CN, whereas inverse scaling emerges in HKP and, at the largest sizes, in MECO. Subsequently, we tested one possible explanation for the inverse scaling in HKP and found that checkpoints whose surprisal remains closer to

n

-gram statistics are better predictors of reading. All in all, the predictive power of surprisal on Chinese reading time measurements is corpus-specific, which cautions against drawing scaling conclusions from a single corpus.

Hongao Zhu, Muxiaoqiao Xu, Yikang Liu +4

Department of Linguistics, UC San Diego · School of Foreign Languages, Shanghai Jiao Tong University · School of Computer Science, Shanghai Jiao Tong University +3

Language Model Scaling Laws Language Modeling Computational Psycholinguistics Surprisal Theory

Ask PDF

Oct 13, 2026

Estimating Surprisal from the Speech Signal

Although language is predominantly spoken, empirical tests of surprisal theory rely almost entirely on written text. In practice, surprisal is estimated using language models (LMs) trained on text and applied to transcriptions, creating a text bottleneck that limits evaluation of the theory in the spoken modality. We remove this bottleneck by estimating surprisal directly from the speech signal using a speech LM. This raises a key question: should surprisal be defined over realization-specific acoustic inputs or over abstract linguistic units? Since a single unit can have many distinct acoustic realizations, each assigned a different probability by the model, defining surprisal over abstract units is difficult. To address this, we introduce an importance sampling estimator that approximates the probability of an abstract unit by marginalizing over realizations generated by a text-to-speech model. Using pupillometry data from a listening study, we evaluate surprisal theory in this text-free setting at multiple levels of granularity. Obtaining surprisal from several speech LMs, we consistently find that surprisal estimated via importance sampling better predicts pupil dilation than realization-specific surprisal for words. We also find that surprisal values for sub-lexical units are predictive of pupil dilation. While we find that surprisal can be estimated without recourse to text, and that marginalizing over distinct realizations of words is more predictive of lexical processing effort, surprisal as a whole is not a strong predictor of change in pupil size, regardless of its source.

Zachary Hopton, Samuel Kiegeland, Francesco Ignazio Re +4

University of Zurich · ETH Zürich · CHI-FRO +1

Surprisal Theory Computational Psycholinguistics Speech Language Models

Ask PDF

Abstract

Surprisal Theory

Explore similar work

The Frequency Confound in Language-Model Surprisal and Metaphor Novelty

A Systematic Analysis of the Predictive Power of LM Surprisal in Reading Chinese

Estimating Surprisal from the Speech Signal