Factors affecting visual word recognition
المؤلف:
Paul Warren
المصدر:
Introducing Psycholinguistics
الجزء والصفحة:
P140
2025-11-08
38
Factors affecting visual word recognition
There are a number of factors that influence both spoken and visual word recognition. These include lexical frequency or the familiarity that participants have with a word and context effects the more constraining the context, the faster and more reliable the recognition of a word. Since these are discussed in Chapter 8, little more will be said about them here, except insofar as they interact with other factors to be discussed in this section. As more is discovered about the process of visual word recognition, so more and more factors have been identified that seem to exert some influence on the process. In this section, we review some of these factors, both because they are in themselves of great interest, but also to provide some background information that will be of relevance in the subsequent discussion of word recognition in the context of reading.
Word superiority effect
The word superiority effect was introduced in the discussion of visual perception for language in Chapter 7. The effect was noted a long time ago (Cattell, 1886). Not only is it easier to identify words than nonwords, but the recognition of letters within a string of letters is easier and more accurate if the string constitutes an existing word. This is reflected in response times and error rates in a range of tasks. So recognition of the letter d is faster if it is contained within a word such as <word> than if it is in a legal nonword e.g. <word> in English or an illegal nonword <wlod>. The word superiority effect indicates that we do not simply process a word letter-by-letter, but rather that recognising a word makes the letters in that word more readily available. That is, there is some top-down processing pathway from words to their constituent letters (McClelland & Johnston, 1977).
Word length effect
A pervasive effect in word recognition is that the size of the word, measured in terms of its length in letters, affects participants’ memories for words they have seen (Baddeley, Thomson & Buchanan, 1975) as well as recognition times for individual words. It has been claimed that longer words take longer to recognise, as reflected in both lexical decision and naming tasks (O’Regan & Jacobs, 1992), although at least one study found this effect for naming but not for lexical decision (Frederiksen & Kroll, 1976). A more recent study was based on responses to over 33,000 words collected in a collaborative project involving six American universities, and with an average of 29 responses to each word New, Ferrand, Pallier Brysbaert, 2006. The data revealed a -shaped pattern of responses – increases in word length between 3 and 5 letters resulted in faster responses, there was no effect of changes from 5 to 8 letters, and responses to words from 8 to 13 letters were increasingly slow. The authors suggest that this pattern may go some way to explaining the somewhat mixed results previously reported, with some authors failing to find a length effect in lexical decision tasks. While the general finding of a relationship between word length and the processing of and/or memory for words is not in itself very surprising, it is nevertheless an effect that needs to be remembered, because research that plans to explore other effects such as those outlined below needs to control for word length in the experimental materials used.
Frequency effect
The frequency effect was discussed in Chapter 3 with respect to lexical access during language production, and in Chapter 8 in connection with spoken word recognition. High-frequency words are recognised more easily and more reliably than low-frequency words. This is reflected in response times and error rates in tasks like lexical decision. It is also the case for English at least that high-frequency words tend to be shorter than low-frequency words. This is known as iZipf ’s law, after one of the early researchers to identify this relationship between word length and frequency Zipf, 1935. Clearly, this is one area where it is particularly important to control for word length when devising experimental materials. So in a study of frequency effects dog should be compared with <ewe> , rather than with <elephant> , since <elephant> is not only less frequent than <dog> but also longer, and if we compared response times to <dog> and <elephant> it would be unclear how much of any effect we observe is due to frequency or length.
Regularity effect
The regularity effect in visual word recognition concerns the relationships between spelling and pronunciation. A regular spelling–sound correspondence means that a word follows the general rules for arriving at a pronunciation based on the spelling. So the <ea> spelling in English most usually has an /i/ pronunciation, so meat is a regular word, while threat is not. The <ave> spelling usually has an /æiv/ pronunciation, so <save> is regular and <have> is not. The basic regularity effect Baron Strawson, 1976 is that visually presented words with a regular spelling sound correspondence are easier to process than those with an irregular correspondence. However, the regularity effect and the frequency effect interact. High-frequency words are not particularly affected by regularity, while low-frequency ones are strongly affected by regularity. We will see later that this is an important finding for advocates of the dual-route model of reading.
Once researchers started looking in more detail at regularity in English, they found that the regularity effect is more complex than previously thought. In particular, they found that consistency of spelling pronunciation correspondences is also important (Patterson & Morton, 1985). That is, visual word recognition is influenced not just by whether the spelling–sound correspondences reflect a set of rules’, but also by the language user’s knowledge of similar lexical entries. The examples in Table 9.1 illustrate the types of consistency and inconsistency that can exist in spelling–sound relationships.
What is particularly important is that there are sets of words, described as gangs’ in Table 9.1, where there is an irregular spelling–sound correspondence for almost all words with that spelling. The example of <look> in the table illustrates this nicely. Most spellings with <o o> have an /u/ pronunciation, such as <soon>, <root>, <scoop>, and so on. The <o ok> body is different. With the exception of <spook> and the rather infrequent <snook> and <dook> known as heroes’ because of how they defy the gang and obey the rules, all <ook> words have the / ʊ / vowel. If we were there fore to encounter a new word e.g. a name coined for a new product, such as <wook>, we would expect the pronunciation to follow the pattern of the gang, rather than that of the heroes. That is, the pronunciation would be consistent, but in terms of the overall pattern of the language, it would be considered irregular because of the general tendency of the <oo> spelling to have an /u/ pronunciation.
Neighbourhood effects
The notion of the neighbourhood was introduced in Chapter 8 in the discussion of competition effects between the activated words in a cohort. In the context of visual word recognition, neighbours are words with similar spelling patterns to the target word. One definition is that these are words sharing all but one letter with the target word, and where letter position is the same (Coltheart, Davelaar, Jonasson & Besner, 1977). So <work>, <ward> and <ford> would all be orthographic neighbours of <word>. A number of properties of the neighbourhood are important for visual word recognition, the most important of which are the neighbourhood size or density, and the lexical frequency characteristics of the neighbourhood, as well as of the target itself. These properties interact with one another, and their effects seem to be to some extent task-dependent (Andrews, 1997; Perea & Rosa, 2000).
A general finding is that responses to low-frequency words, but not those to high-frequency words, are affected by neighbourhood size. The nature of this effect for low-frequency words is not the same in all tasks. First, responses in naming tasks i.e. where the word is read aloud are generally faster for low-frequency words that come from large or dense neighbourhoods such rink as, which has 15 neighbours, including (sink, rank, risk, ring) than for low-frequency words from small or sparse neigh bourhoods (e.g. tact, with 6 neighbours Andrews, 1989). This shows a facilitatory effect of neighbourhood size. Second, inhibitory effects have emerged in tasks which require the unique identification of a target word. Such a task is progressi e demasking, where more and more of a word is exposed on repeated presentations until the participant is able to uniquely identify the target. The more neighbours there are for a word, the longer the exposure to a word needs to be before it can be recognised (Carreiras, Perea & Grainger, 1997). Third, lexical decision tasks have variably shown facilitation or inhibition from neighbourhood size, depending on other aspects of the task, including the degree of word-likeness of the nonwords used in the task (Andrews, 1992), an emphasis of speed over accuracy (Grainger & Jacobs, 1996), and the language being studied. A further effect is that the frequency characteristics of words in the neighbourhood also have an impact on recognition – words with high-frequency neighbours are recognised more slowly than comparable words with low-frequency neighbours (Pollatsek, Perea & Binder, 1999).
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
