In contrast to the item based explanation the list-based explanation of the word length effect has also been supported by a memory model. The Working Memory model (Baddeley, 1986) ascribes the word-length effect to decay offset by rehearsal. Items are stored in a phonological store, which is sensitive to time-based decay. However rehearsal can refresh or reactivate the traces in the store to counter the effects of decay, even if it were a covert process. "The amount of verbal information that can be retained is a trade-off between the decay rate (which is assumed to be fixed) and the covert articulation rate, which can vary". A number of studies have been carried out in which researchers have computed span and measured articulation and have then solved for the duration of the verbal trace. This value is has been generally found to always be around 1.5 to 2 seconds (Schweickert & Boruff, 1986). Therefore using this model to explain the word length effect shows that rehearsal rate depends upon all items in a list and whether a particular item is recalled depends on the other items in the list.
As the word length effect is linked with the operation of the articulatory control process used to maintain the contents of the store. Articulatory rehearsal appears to take place in real time it has thus been argued that lists containing items that take longer to articulate will receive less rehearsal, and as a result will be lost from the articulatory loop. This interpretation of the word length effect is supported with evidence from linear relationships between the number of words that can be recalled and the rate at which they are uttered, (Baddeley et al. 1975).
After looking at both types of explanations given to the word length effect, is the crucial feature the length of time it takes to utter the items or the number of syllables. A number of studies have shown that it is the duration, which appears to be the critical factor. In these studies it has been found that words which have long vowels and that are spoken slowly lead to shorter spans than words with the same number of syllables and phonemes that are spoken more rapidly, suggesting that time is the crucial factor. The following studies should shed some light on whether there is really a time based word length effect.
Naveh- Benjamin and Ayres (1986) measured relationship between reading time and the memory span in English, Spanish Hebrew and Arabic. They concluded that faster speeded and normal- pace reading rates for language were associated with larger memory span for speakers of that language. Hence Baddeley felt that there must be a strong correlation between the rate at which a person speaks and their memory span (Baddeley et al 1975). The Following studies appear to support Baddeleys assumption.
Ellis and Hennelly (1980) used the word length effect to interpret the findings of an intelligence test used with Welsh speaking children. These children appeared to have a consistently lower performance on the digit span subtest in comparison to their English contemporaries. Ellis and Hennelly proposed that the digit names used in the Welsh language even though they have the same number of syllables as in the English language tend to have longer vowel sounds and take longer to say. They tested this hypothesis using bilingual welsh and English speakers. The participants were found to have poorer spans in their native language of Welsh than in English, but as predicted their spans were equal when measured in terms of spoken time. However when these participants were prevented from rehearsing by being asked to rehearse irrelevant sounds the difference between span in the two languages disappeared.
The question arises here of whether the effect was limited purely to memory span? Ellis and Hennelly don’t feel that this was the case since their subjects also showed some signs of slower performance and higher error rate in mental arithmetic using Welsh digits. These finding has also been supported by the work of Naveh et al in a various of languages obtaining a clear relationship between memory span and the time it takes to articulate the digits 1-10 in that particular language.
Another study attempted to find whether the crucial factor was the duration of the spoken word or its length in terms of syllables. The experiment compared immediate memory for two sets of words. Both lists comprised of two syllables words but one set was spoken relatively briefly compared to the other. Baddeley (1975) observed a consistent tendency for subjects to do better on the short duration words even though the two sets were matched for syllables and phonemes. Thus indicating that the time taken to articulate (duration) was the crucial variable and not the word type. This finding is also consistent with the trace decay hypothesis which suggests that duration is important as longer words take longer to say and hence allow the memory trace to be refreshed less frequently. These findings are also based on the most widely accepted view of immediate memory, which is the framework known as working memory (Baddeley, 1986; Cowan, 1995; Miyake & Shah, 1999).
Despite a great number of studies supporting a time based word length effect, there are also studies and experiments in opposition to this explanation. The second type of explanation does not support the idea that the effect is based on time. To clarify this three experiments were carried out to compare recall of three different sets of disyllabic words that differed systematically only in spoken duration (Ian Neath, Tamra J. Bireta B and Aim M. Surprenant). One set of words showed a word length effect, the second set showed no effect of word length, and the third showed a reverse word length effect, with long words recalled better than short. A fourth set of words was then created, which also failed to show a time-based word length effect. As all four experiments used the same methodology and only varied in the spoken duration, it is argued that the time-based word length effect is not as powerful as it may be assumed. These findings pose problems for models based on the phonological loop (the working memory model).
The works of Lovatt, Avons, and Masterson (2000) also presented data suggesting that the time-based word length effect is observable only with one set of stimuli: other stimulus sets produce no such effect, or even a reverse effect (as seen in the above experiments). Studies that have used different stimulus sets have continually failed to replicate the time-based word length effect. Caplan, Rochon, and Waters (1992) created a set of two-syllable words that were matched for number of syllables and number of phonemes but differed in spoken duration. They observed better recall of the long words. Similar results have also been reported in additional languages.
Zhang and Feng (1990; reported in Lovatt et al., 2000) also found no difference in the level of recall of short and long Chinese disyllables.
In addition to this a number of experiments have been carried out to test the word length effect in both pure lists and mixed lists of words. In an experiment carried out by Hulme et al they examined immediate serial recall of six-item lists with spoken presentation and spoken recall. The procedure corresponds to that most commonly used in studies of the word-length .To ensure that clear effects of word-length will be obtained they compared the recall of words that differed greatly in length (1-syllable and 5-syllable words). The recall of pure lists of long words and pure lists of short words, were compared with recall of alternating lists of long and short words. They found in the pure lists of long and short words show a large advantage for the recall of short words, replicating earlier findings. In the alternating lists the recall of short and long words was at similar levels to the recall of short words in pure lists. In a second experiment constructed by Hulme et al in which they presented both alternating and randomly arranged mixed lists, and switched to visual rather than auditory presentation. They found that there was a strong word-length effect in the pure conditions, but this was not the case in the alternating and random conditions, where the long and short words were recalled as well as short words in the pure lists.
Even though the above studies used a combination of different types of sets they still appear to show that the time based word length effect is subject to specific stimulus sets. The question also arises regarding the loss of the word length effect in mixed lists. If the word length effect is based on the amount of time it takes to articulate the stimuli then why is the effect lost when the words are put in alternating lists as this does not effect the duration. Despite there being strong support for the time based word length effect these flaws seem to limit its credibility. If these flaws are seen as showing that that there is no word length effect this could have immense theoretical significance. As the basic structure of the working memory framework
(Baddeley, 1986, 2000) consists of the phonological store and articulatory control process, which together make up the articulatory loop. As we discussed before if the articulatory loop process does not rehearse items then items will decay, and because shorter items take less time to rehearse, more traces of short items can be refreshed than traces of long items, which was the basic explanation of the time based word length effect by the list-based explanation. However if the time based word length effect is not found in different stimulus sets, then the very foundation of the working memory view is "greatly compromised". And taking into consideration that the working memory model is the most widely accepted view of immediate memory; this has even greater effects on the research and theories of memory as a whole. Despite the time based word length effect appearing as “common sense” as it seems quite clear that things which take longer to say or read with be harder to remember (due to the time it takes). In order for the time based word effect to prove it strength it will have to be apparent not only in different languages but also in different stimulus sets.