READING: BOTTOM-UP VS TOP-DOWN
المؤلف:
John Field
المصدر:
Psycholinguistics
الجزء والصفحة:
P232
2025-10-01
327
READING: BOTTOM-UP VS TOP-DOWN
Goodman (1967) put forward what is sometimes termed a ‘top-down’ approach to reading, describing it as a ‘psycholinguistic guessing game’. He suggested that skilled reading was marked by the reader’s ability to use contextual information in order to anticipate the words which came next. This enabled the good reader to reduce his/her dependence upon decoding symbols on the page. Word recognition could take place more rapidly and the skilled reader could even skip words that were highly predictable. Frank Smith (1971) took a similar view, arguing that skilled readers exploit redundancy in a text. He claimed that the speed of skilled reading showed that the reader could not attend to every single letter; instead, they must make use of syntactic and semantic information to increase their efficiency. He even argued that, on this basis, it was futile to teach decoding skills to children; one should simply encourage them to ‘make sense’ of texts.
The ‘top-down’ view does not accord with current theory about the memory processes involved in reading. Decoding is conceived as being highly automatic, which means that it makes small demands upon working memory. By contrast, word or content prediction of the kind postulated by Goodman is under the control of the reader, and would thus be far more demanding (and slower and less efficient) than the process which it supposedly replaces. In addition, much of a reader’s working memory appears to be committed to the difficult tasks of integrating incoming information into the mental representation of the text that has been constructed so far and of monitoring for failures of understanding. These are much more pressing demands upon limited resources than the need to predict what comes next.
Goodman’s theory has been influential, but there is no body of evidence that supports it. Proponents of the theory claim that results from miscue analysis show that a large proportion of reading aloud errors are influenced by context; but this finding has been challenged. Considerable evidence suggests that skilled readers do not anticipate words as Goodman suggested:
Eye movement data shows only a small decrease in fixation time when a word is predictable from the context. The conclusion is that there is no point in compromising the efficiency of the highly automatic process of eye movement by incorporating a more conscious (and therefore slower) context-driven process.
Readers show no evidence of making explicit forward guesses. Indeed, with a text that is authentic and not specially written, they are extremely inaccurate when asked to predict forthcoming words.
Poor readers have been shown to be more sensitive to context than good.
Only about 40 per cent of function words and 10 per cent of content words can be predicted from context.
The most widely accepted view is that it is efficient decoding which makes for skilled reading. Perfetti’s verbal efficiency theory (or bottle-neck hypothesis) suggests that the processes of decoding and comprehension compete for limited space in working memory. Where decoding is slow, it results in smaller amounts of information being made available at any time, and therefore a focus upon local rather than global meaning relations. Where it is not automatic, it demands extra working memory capacity, leaving less for other processes. In addition, the contents of working memory decay rapidly– so information derived from a slow decoding process may be lost before it can be analysed.
There is evidence that training in rapid decoding does not improve the comprehension of weak L1 readers. What they appear to need is accurate and automatic decoding– not a higher reading speed alone. However, inaccurate decoding produces the same outcome as slow decoding. If a weak reader has to keep regressing to check word recognition, the result is to slow down the supply of data and hence to encourage a focus on small-scale rather than larger-scale patterns of meaning.
Investigating the ‘top-down’ hypothesis, Stanovich reviewed 22 studies of reading and found no evidence that good readers used context to support word recognition as Goodman had suggested. There was evidence that skilled readers do make greater use of context (mental model of the text so far, world knowledge, text schemas) but it is in order to enrich meaning. Stanovich argued that context is used in two distinct ways:
to construct global meaning and to self-monitor;
to compensate for inadequate decoding skills.
The first is more characteristic of the skilled reader and the second of the unskilled. Stanovich’s interactive compensatory hypothesis expands on the second function. It envisages a trade-off between message quality and the extent to which top-down information is relied upon to support decoding. If a reader has poor decoding skills, then they may fall back on top-down information. However, if a text is degraded, then even a skilled reader may have recourse to top-down information to compensate for gaps in the message.
See also: Bottom-up processing, Eye movements, Reading: decoding, Reading speed, Top-down processing
Further reading: Gough and Wren (1999); Oakhill and Garnham (1988); Perfetti (1985: Chaps 6–7); Rayner and Pollatsek (1989); Stanovich (1980)
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
