INTERACTIVE ACTIVATION
المؤلف:
John Field
المصدر:
Psycholinguistics
الجزء والصفحة:
P137
2025-09-03
339
INTERACTIVE ACTIVATION
A model of language processing which assumes that all levels of representation freely influence each other. This might mean, for example, that a listener’s perception of a distorted sound at the beginning of the sequence eel would be influenced at the time of processing by knowledge of the word WHEEL or by a context such as ‘He changed the eel on the car’. An interactive account contrasts with modular accounts which suggest that the sequence eel must first be processed phonologically; only then can information from other levels be brought in to disambiguate it.
Interactive activation models thus present the processing of a spoken or written word as subject simultaneously to both bottom up (data driven) and top-down (context driven) influences. Their proponents argue that this makes all sources of information immediately available to the listener or reader, enabling an informed choice to be made as to the identity of the word. Their opponents argue that a model of this kind overloads the processor with information, making a decision more difficult.
The various cues are said to assist recognition by supplying activation to a particular letter or word. Thus, evidence of WHI on the page would provide strong bottom-up activation for the words WHICH and WHITE but only weak activation for partial matches like WHO or WRITE. If top-down information suggested that a question word was needed, WHICH would receive further activation. The activation of a letter or word can be increased by a positive (facilitatory) connection or reduced by a negative (inhibitory) one. When one particular word is supported by overwhelming evidence, it ‘fires’, i.e. becomes accepted as the correct match.
A distinction can be made between interactive models which incorporate between-levels activity and those which also feature within levels activity. In a within-levels model, activation supporting the word WHICHhas an inhibitory effect on other likely words– reducing the activation of competitors such as WHITE, WRITE etc. Similarly, at letter level, any increase in the activation of H would reduce the activation of R.
Although the term ‘interactive activation’ is now used generically for this type of account of listening or reading, it was originally used for a specific connectionist model of visual word recognition (McClelland and Rumelhardt, 1981). This model operates on three levels: feature, letter and word, with excitatory and inhibitory connections between elements at all levels. Recognition of the word WORD is supported, bottom-up, by evidence from features (oblique strokes, circle etc.) and letters; but it is also supported, top-down, by the existence of WORD in the lexicon. In this way, the model accounts neatly for word superiority effects, where letters prove easier to detect when they occur in actual words than when they occur in non words or even when they occur alone.
Similar interactive models (TRACE, Shortlist etc.) have been designed to account for auditory perception.
See also: Activation, Bottom-up processing, Modularity2, Top-down processing
Further reading: Carroll (1999: 97–9); Whitney (1998: 187–9)
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
