Frequency, competition and neighbourhoods
المؤلف:
Paul Warren
المصدر:
Introducing Psycholinguistics
الجزء والصفحة:
P132
2025-11-06
29
Frequency, competition and neighbourhoods
The earlier illustration of the activation of word candidates needs an important supplement, since a regular and very robust research finding is that there are freuency effects in word recognition (Forster Chambers, 1973; Monsell, 1991; Murray & Forster, 2004). That is, words that we encounter more often have an advantage over words that we do not see or hear so often. This is reflected in faster response times and greater accuracy for common words in tasks such as lexical decision. In Figure 8.2, this frequency effect is shown by a slower increase in activation for captive and capital relative to captain, so that even when the phonetic information still matches both captain and captive (e.g. at /kæpt/), the former already shows an activation advantage. This advantage is also reflected in the detailed results of the cross-modal priming experiments referred to earlier – more frequent word candidates showed greater priming effects than less frequent candidates.
The illustration in Figure 8.2 suggests that frequency effects might be accounted for in terms of the rate of increase of activation for the different word candidates. Alternative mechanisms have been suggested, including different resting levels of activation for different words, so that frequent words have an initial advantage by having a higher resting level than rare words. Under such a mechanism, if the same input e.g. the /kæpt/ onset in our illustration results in the same rate-of-change in activation of word candidates, the resting-level advantage will persist for more frequent words, so that these cross some activation threshold and achieve recognition before the less frequent words. An effect that is related to the frequency effect, and which can probably be accounted for in a similar manner, is that of recency Oliphant, 1983. That is, words which we have encountered quite recently seem to gain – at least temporarily – an advantage over similar words that have not been heard for a while. If using a word causes an increase in its activation, and if it takes a while for this activation to return to its resting level, then hearing or seeing the word a second time may take place at a point where the activation is still above its normal resting level, resulting in faster recognition. In neurophysiological studies, it has been demonstrated that frequency effects and those of other factors discussed in this chapter and the next are reflected in the size of the electrical response to words in ERP studies, but also in the speed of certain components of this response (see Kutas Federmeier, 2007). Both of these findings indicate greater neural stimulation in response to higher frequency words.
Competition
The recognition of a word depends quite clearly not only on the properties of that word itself its sound form, its frequency, and so on but also on the properties of other words that compete with it. We saw earlier that the use of nasalisation information on a vowel in an English word like soon helps listeners to distinguish this word from similar-sounding words such as suit or soup. This means that given the input /sữ/ above the vowel indicates nasalisation, listeners may already be forming a strong hypothesis that the word is soon and not suit or soup. This is reflected in the results of gating experiments with words like this. However, the same nasalisation information has a different information value if it occurs on the vowel in, since there is no word that has a nasal consonant after the /u/ vowel here. If it tells us anything, nasalisation here would probably indicate that the speaker is one who typically has a nasal setting to their speech, i.e. it might help us identify the speaker, but not the word. What this boils down to is that the informativeness of properties of the signal depends on the alternative possibilities that exist and whether these alter natives can be ruled in or out by these properties. This is the notion of contingency of choice – knowing you have heard the word cat depends not just on the sounds of cat, but also on knowing that you have not heard the words cap, can, cash, etc. (Warren & Marslen-Wilson, 1987, 1988).
Such considerations make it obvious that we need to take into account not just individual words, but also the relationships of these words to their competitors. The notion of competition is an important one in models of word recognition, and has been studied in considerable depth. Like frequency effects, competitor effects have been modelled in different ways by different researchers. One of the most popular notions is that a result of competition between two words is that the words inhibit one another. That is, competition has a negative impact on the activation levels of competitor words. The size of this inhibition effect might depend on the relative activation levels of the words that are in competition with one another. Given an ambiguous input like /kæpt/, for instance, the more frequent word captain will have a higher activation level than the less frequent captive, and this greater activation level will allow captain to exert a stronger inhibitory influence on a e than vice versa. This will in effect boost the activation difference between the two words, making it more certain that captain will be recognised, in the absence of any further differentiating information.
Neighbourhoods
Another important aspect of competition is the size of the competitor set. A word belongs to a neighbourhood of words that share similar proper ties (Luce & Pisoni, 1998; Pisoni, Nusbaum, Luce Vitevitch, Stamer Slowiaczek, 1985; Sereno, 2008). A word-initial cohort is of course a type of neighbourhood, but more extensive neighbourhoods than these have been examined, for instance neighbourhoods that consist of all words that are just one phoneme different from the target word under consideration, regardless of where in the word the differentiating phoneme is see example in sidebar. The recognition of a word in tasks such as lexical decision and naming, i.e. reading aloud is affected by how many neigh bours it has, by the phonetic similarity of these words to the target word i.e. the extent of the overlap of the forms of the words, and by the frequencies of the words in the neighbourhood. For instance, a word with a lot of high-frequency neighbours is identified more slowly and less accurately than a word with only a few neighbours or with only low-frequency neighbours. These factors contribute to what has become known as the neighbourhood density of a word, which has been shown to have an influence on both word recognition influence on both word recognition (Vitevitch & Luce, 1998) and word production (Vitevitch, 2002).
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
