TY - JOUR
T1 - Repetition Effects in Word and Pseudoword Identification. Comment on Salasoo, Shiffrin, and Feustel
AU - Johnston, James C.
AU - van Santen, Jan P.H.
AU - Hale, Benita L.
PY - 1985/12
Y1 - 1985/12
N2 - Two recent articles by Feustel, Shiffrin, and Salasoo (1983) and Salasoo, Shiffrin, and Feustel (1985) argue that word identification is based on episodic memory as well as semantic (or "permanent, abstract") memory. The earlier article argued for separate processing stages affected by repetition (episodic memory) and lexicality (semantic memory). To account for the new finding that number of repetitions interacts with lexicality, the later article invokes the same two types of memory, operating in parallel rather than serially. We argue that Salasoo et al.'s data are compatible with a wide variety of competing theories, including some that do not involve episodic memory at all. We question the relevance of a complex formal model shown by the authors to account for the main trends in the data, because it makes little use of key properties of episodic memory. Furthermore, we show that two variants of a less complex model inspired by logogen theory fit the data as well or better than the model of Salasoo et al., using fewer free parameters. We argue that what is needed is not an existence proof that one particular complex model can fit a body of data, but experimental manipulations that successfully discriminate between broad classes of competing theories.
AB - Two recent articles by Feustel, Shiffrin, and Salasoo (1983) and Salasoo, Shiffrin, and Feustel (1985) argue that word identification is based on episodic memory as well as semantic (or "permanent, abstract") memory. The earlier article argued for separate processing stages affected by repetition (episodic memory) and lexicality (semantic memory). To account for the new finding that number of repetitions interacts with lexicality, the later article invokes the same two types of memory, operating in parallel rather than serially. We argue that Salasoo et al.'s data are compatible with a wide variety of competing theories, including some that do not involve episodic memory at all. We question the relevance of a complex formal model shown by the authors to account for the main trends in the data, because it makes little use of key properties of episodic memory. Furthermore, we show that two variants of a less complex model inspired by logogen theory fit the data as well or better than the model of Salasoo et al., using fewer free parameters. We argue that what is needed is not an existence proof that one particular complex model can fit a body of data, but experimental manipulations that successfully discriminate between broad classes of competing theories.
UR - http://www.scopus.com/inward/record.url?scp=0022336313&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0022336313&partnerID=8YFLogxK
U2 - 10.1037/0096-3445.114.4.498
DO - 10.1037/0096-3445.114.4.498
M3 - Article
C2 - 2934500
AN - SCOPUS:0022336313
SN - 0096-3445
VL - 114
SP - 498
EP - 508
JO - Journal of Experimental Psychology: General
JF - Journal of Experimental Psychology: General
IS - 4
ER -