Linguistic prediction is a phenomenon in psycholinguistics occurring whenever information about a word or other linguistic unit is activated before that unit is actually encountered. Evidence from eyetracking, event-related potentials, and other experimental methods indicates that in addition to integrating each subsequent word into the context formed by previously encountered words, language users may, under certain conditions, try to predict upcoming words. In particular, prediction seems to occur regularly when the context of a sentence greatly limits the possible words that have not yet been revealed. For instance, a person listening to a sentence like, "In the summer it is hot, and in the winter it is..." would be highly likely to predict the sentence completion "cold" in advance of actually hearing it. A form of prediction is also thought to occur in some types of lexical priming, a phenomenon whereby a word becomes easier to process if it is preceded by a related word.Neely, J. (1991). Semantic priming in visual word recognition: a selective review of current theories and findings. In: Basic processes in reading: visual word recognition (Besner D, Humphreys GW, eds), pp 264 –336. Hillsdale, NJ: Erlbaum Associates. Linguistic prediction is an active area of research in psycholinguistics and cognitive neuroscience.
Evidence from eyetracking
Visual world paradigms In the eyetracking visual world paradigm, experimental subjects listen to a sentence while staring at an array of pictures on a computer monitor. Their eye movements are recorded, allowing the experimenter to understand how language influences eye movements toward pictures related to the content of the sentence. Experiments of this type have shown that while listening to the verb in a sentence, comprehenders anticipatorily move their eyes to the picture of the verb's likely direct object (e.g. "cake" rather than "ball" while hearing, "The boy will eat...").Altmann, G. & Kamide, Y. (1999). Incremental interpretation at verbs: restricting the domain of subsequent reference. Cognition, 73, 247–264. Subsequent investigations using the same experimental setup showed that the verb's subject can also determine which object comprehenders anticipate (e.g., comprehenders look at the merry-go-round rather than the motorcycle while hearing, "The little girl will ride...").Kamide, Y., Altmann, G. & Haywood, S. (2003). The time-course of prediction in incremental sentence process- ing: evidence from anticipatory eye movements. Journal of Memory and Language, 49, 133–159. In short, comprehenders use the information in the sentence context to predict the meanings of upcoming words. In these experiments, comprehenders used the verb and its subject to activate information about the verb's direct object before hearing that word. However, another experiment has shown that in a language with more flexible word order (German), comprehenders can also use context to predict the sentence's subject.Kamide, Y., Scheepers, C. & Altmann, G. (2003). Integration of Syntactic and Semantic Information in Predictive Processing: Cross-Linguistic Evidence from German and English. Journal of Psycholinguistic Research, 32 (1), 37-55.
Natural reading Eyetracking technology has also been used to monitor readers' eye movements while they read text on a computer screen. Data from this kind of experiment has supported the hypothesis that readers use contextual information to predict upcoming words during natural reading. Specifically, readers fixate their eyes on a word for a shorter time when the word occurs in a moderately or highly constraining context, compared to the same word in an unconstrained context. This is true regardless of the word's frequency or length. Readers are also more likely to skip over a word in a highly constraining context only.Rayner, K. & Well, A. (1996). Effects of contextual constraint on eye movements in reading: A further examination. Psychonomic Bulletin and Review, 3 (4), 504-509. Subsequent investigations of reading in the Chinese logographic script have shown that despite the large differences between the Chinese and English orthographies, readers exploit contextual information for prediction in similar ways, with the exception that Chinese readers were more likely to skip words in moderately constraining contexts.Rayner, K., Li, X., Juhasz, B. & Yan, G. (2005). The effect of word predictability on the eye movements of Chinese readers. Psychonomic Bulletin and Review, 12 (6), 1089-1093.
Computational models of eye movements during reading, which model data related to word predictability, include Reichle and colleagues' E-Z Reader modelReichle, E., Rayner, K. & Pollatsek, A. (2003). The E-Z Reader model of eye movement control in reading: Comparisons to other models. Behavioral and Brain Sciences, 26, 445-526. and Engbert and colleagues' SWIFT model.Engbert, R., Nuthmann, A., Richter, E. & Kliegl, R. (2005). SWIFT: A dynamical model of saccade generation during reading. Psychological Review, 112 (4), 777-813.
Evidence from event-related potentials
M100 The M100 discussed here is the magnetic equivalent of the visual N1 potential—an event-related potential linked to visual processing and attention. The M100 was also linked to prediction in language comprehension in a series of event-related magnetoencephalography (MEG) experiments. In these experiments, participants read words whose visual forms were either predictable or unpredictable based on prior linguistic contextDikker, S., Rabagliati, H., Farmer, T. & Pylkkänen, L. (2010). Early occipital sensitivity to syntactic category is based on form typicality. Psychological Science, 21(5), 629-634.Dikker, S., Rabagliati, H. & Pylkkänen, L. (2009). Sensitivity to syntax in visual cortex. Cognition, 110 (3), 293-321. or based on a recently seen picture.Dikker, S. & Pylkkänen, L. (2011). Before the N400: Effects of lexical–semantic violations in visual cortex. Brain and Language 118, 23-28. The predictability of the word's visual form (but not the predictability of its meaning) affected the amplitude of the M100. There is ongoing controversy about whether this M100 effect is related to the early left anterior negativity (eLAN), an event-related potential response to words that is theorized to reflect the brain's assignment of local phrase structure.Friederici, A. & Weissenborn, J. (2007). Mapping sentence form onto meaning: The syntax-semantic interface. Brain Research, 1146, 50-58.
P2 The P2 component is generally thought to reflect higher-order perceptual processing and its modulation by attention. However, it has also been linked to prediction of visual word forms. The P2 response to words in highly constraining contexts is often larger than the P2 response to words in less constraining contexts. When experimental participants read words that are presented to the left or right of their visual fixation (stimulating the opposite hemisphere of the brain first), the larger P2 for words in highly constraining contexts is observed only for right visual field presentation (targeting left hemisphere).Wlotko, E. & Federmeier, K. (2007). Finding the right word: Hemispheric asymmetries in the use of sentence context information. Neuropsychologia, 45, 3001-3014. This is consistent with the PARLO hypothesis that linguistic prediction is mainly a function of the left hemisphere, discussed below.
N400 The N400 is part of the normal ERP response to potentially meaningful stimuli, whose amplitude is inversely correlated with the predictability of a stimuli in a particular context.Kutas, M., & Hillyard, S. A., (1984). Brain Potentials during reading reflect word expectancy and semantic association. Nature, 307, 161-163. In sentence processing, the predictability of a word is established by two related factors: 'cloze probability' and 'sentential constraint'. Cloze probability reflects the expectancy of a target word given the context of the sentence, which is determined by the percentage of individuals who supply the word when completing a sentence whose final word is missing. Kutas and colleagues found that the N400 to sentences final words with cloze probability of 90% was smaller (i.e., more positive) than the N400 for words with cloze probability of 70%, which was then smaller for words with cloze probability of 30%. Closely related, sentential constraint reflects the degree to which the context of the sentence constrains the number of acceptable continuations. Whereas cloze probability is the percent of individuals who choose a particular word, constraint is the number of different words chosen by a representative sample of individuals. Although words that are not predicted elicit a larger N400, the N400 to unpredicted words that are semantically related to the predicted word elicit a smaller N400 than when the unpredicted words are semantically unrelated. When the sentence context is highly constraining, semantically related words receive further facilitation in that the N400 to semantically related words is smaller in high constraint sentences than in low constraint sentences.Federmeier, K. D., & Kutas, M. (1999). Right words and left words: electrophysiological evidence for hemispheric differences in meaning processing. Cognitive Brain Research, 8, 373-392.Federmeier, K. D., McLennan, D. B., De Ochoa, E., Kutas, M. (2002). The impact of semantic memory organization and sentence context information on spoken language processing by younger and older adults: an ERP study. Psychophysiology, 39, 133-146.Federmeier, K. D. (2007). Thinking ahead: the role and roots of prediction in language comprehension. Psychophysiology, 44, 491-505. Evidence for the prediction of specific words comes from a study by DeLong et al.DeLong, K. A., Urbach, T. P., Kutas, M. (2005). Probabilistic word pre-activation during language comprehension inferred from electrical brain activity. Nature Neuroscience, 8, 1117-1145. DeLong and colleagues took advantage of the use of different indefinite articles, 'A' and 'AN' for English words that begin with a consonant or vowel respectively. They found that when the most probable sentence completion began with a consonant, the N400 was larger for 'AN' than for 'A' and vice versa, suggesting that prediction occurs at both a semantic and lexical level during language processing. (The study never replicated. In the most recent multi-lab attempt (335 participants), no evidence for word form prediction was found (Niewland et al., 2018).
Late positivity
P300
The P300, specifically the P3b is an ERP response to improbable stimuli and is sensitive to the subjective probability that a particular stimulus will occur. The P300 has been closely tied to context updating, which can be initiated by unexpected stimuli.Donchin, E., & Coles, M. G. H., (1988). Is the P300 component a manifestation of context updating? The Behavioral and Brain Sciences, 11, 357–374.
P600