This paper presents a probabilistic Earley parser as a psycholinguistic model for understanding human sentence processing. The parser is used to compute cognitive load, defined as the total probability of structural options that have been disconfirmed during sentence processing. This load is calculated using a probabilistic Earley parser, which is interpreted as generating predictions about reading time on a word-by-word basis. The parser is based on probabilistic context-free grammar (PCFG) and is used to model processing phenomena associated with garden path structural ambiguity and subject/object relative asymmetry. The paper outlines three principles that the model must satisfy: strong competence, frequency affects performance, and sentence processing is eager. Strong competence implies that the human sentence processing mechanism directly uses grammar rules. Frequency affects performance suggests that statistical theories of language performance are valid. Sentence processing is eager means that the parser is not rushed and does not ignore information that could be useful. The probabilistic Earley parser computes prefix probabilities, which are used to calculate cognitive load. These probabilities are derived from the grammar and are used to model the difficulty of parsing sentences. The parser is able to account for garden path effects by showing how the probability of certain structures decreases as new information is processed. The parser also explains subject/object relative asymmetry by showing how the probability of certain structures changes based on the context. The paper concludes that a probabilistic grammar-based model can explain important processing phenomena, including structural ambiguity and subject/object asymmetry. It supports the idea that phrase-level contingent frequencies can account for processing effects previously attributed to other mechanisms. The model provides clear, testable predictions and has the potential to offer new insights into psycholinguistics.This paper presents a probabilistic Earley parser as a psycholinguistic model for understanding human sentence processing. The parser is used to compute cognitive load, defined as the total probability of structural options that have been disconfirmed during sentence processing. This load is calculated using a probabilistic Earley parser, which is interpreted as generating predictions about reading time on a word-by-word basis. The parser is based on probabilistic context-free grammar (PCFG) and is used to model processing phenomena associated with garden path structural ambiguity and subject/object relative asymmetry. The paper outlines three principles that the model must satisfy: strong competence, frequency affects performance, and sentence processing is eager. Strong competence implies that the human sentence processing mechanism directly uses grammar rules. Frequency affects performance suggests that statistical theories of language performance are valid. Sentence processing is eager means that the parser is not rushed and does not ignore information that could be useful. The probabilistic Earley parser computes prefix probabilities, which are used to calculate cognitive load. These probabilities are derived from the grammar and are used to model the difficulty of parsing sentences. The parser is able to account for garden path effects by showing how the probability of certain structures decreases as new information is processed. The parser also explains subject/object relative asymmetry by showing how the probability of certain structures changes based on the context. The paper concludes that a probabilistic grammar-based model can explain important processing phenomena, including structural ambiguity and subject/object asymmetry. It supports the idea that phrase-level contingent frequencies can account for processing effects previously attributed to other mechanisms. The model provides clear, testable predictions and has the potential to offer new insights into psycholinguistics.