Performance in immediate serial recall tasks is determined by type of material (digits, letters, words), word length and the phonological confusability of the words to remember. The phonological similarity effect refers to the impairment of immediate serial recall performance for items that are of high phonological similarity. While research on this issue is based on intuitive estimations of phonological item similarity, our study introduces a metrics for quantifying this similarity. In two Experiments we test the influence of this metrics on the phonological similarity effect for visually presented words. Experiment I shows a gradual phonological similarity effect on transposition errors when item similarity is continuously varied. Experiment II replicates this result using an irrelevant speech manipulation which reduces the phonological similarity effect. A test of Baddeley's (1992) phonological store hypothesis in Experiment II failed to yield positive evidence.
Key words: phonological similarity effect, phonological store hypothesis, changing state hypothesis, irrelevant speech
Research on short-term memory performance shows a major influence of the type of the material to remember (e.g., Baddeley, 1968; 1976; 1986; 1997; Baddeley & Hitch, 1974). For example, memory span (typically the list length recalled correctly in 50% of the time) in immediate serial recall varies with materials such as digits, letters, words, or nonwords (e.g., Baddeley, 1976; Cavanagh, 1972; Conrad, 1964; Jacobs, 1887). One issue in this research is the phonological similarity effect which deals with the result that material which is highly confusable in terms of phonological properties is more difficult to produce from memory than less confusable material (e.g., Baddeley, 1968; Henson, Norris, Page, & Baddeley, 1996; Page & Norris, 1998a). The phonological similarity effect has been interpreted as evidence in favor of one of the major constructs in this field, the phonological loop, a working memory component supposed to hold memory entries with properties common to speech or to sound in a broader sense (Baddeley, 1968). If rehearsal in the phonological loop is suppressed by counting backwards or producing some other kind of concurrent articulatory activity the phonological similarity effect is vanishing (e.g., Murray, 1968; see also Estes, 1973; Levy, 1971; Surprenant, Neatly & LeCompte, 1999).
Typically, the phonological similarity effect occurs as a poorer performance in recall when items are phonologically similar to each other (e.g., the rhyming letter names B, C, D, G) as compared to items that are more phonologically distinct (e.g., the letter names F, X, Z, Q, W). Furthermore, there is evidence that in mixed lists recall of nonconfusable items is insensitive to the degree of confusability of the other list items (Baddeley, 1968; Henson, Norris, Page, & Baddeley, 1996). The experimental markers most widely used to show a phonological similarity effect are error rates. In immediate serial recall errors can be classified into substitutions and omissions. Substitutions occur when an erroneous item is recalled in a given position, omissions occur when no item is recalled. Substitutions can be subdivided into transpositions (an erroneous item was itself presented in another position in the list) and intrusions, which refer to items not presented in the target list (Page & Norris, 1998a). If the error proportion is measured for each serial position of recall a typical serial position curve is obtained. It shows increasing error proportions for increasing serial position plus an advantage for items in the final position.
A second issue in this field of research is the irrelevant speech (sound) effect: When irrelevant or unattended speech or sound is added during the processing of the material, recall performance is usually decreased (e.g., Baddeley, 2000; Baddeley & Salame, 1986; Ellermeier & Hellbruck, 1998; Ellermeier & Zimmer, 1997; Jones, 1999; Jones & Macken, 1995; Jones & Tremblay, 2000; Neath, 2000; Tremblay, Macken, & Jones, 2000). …