By:
Dawn Blasko (dgb6@psu.edu)
Associate Professor of Psychology
Penn State Erie

Back to Instructor's Page

Introduction

The lexical decision experiment is useful to address several issues in the areas of psycholinguistics, memory and research methods. Many models of the organization of memory posit a network of links between semantically related concepts. Most assume that when we activate one concept in memory the activation spreads to related associated links. This spreading activation account has been used to explain the finding that a concept can often be “primed” by related information. In other words, it is generally faster to recognize one concept if a related concept has first been encountered. In this experiment the lexical decision task of Meyer and Schvaneveldt (1971) is replicated and extended. In their experiment two letter strings were presented and the subjects made a “lexical decision.” That is, they decided if “both” strings were real words. Some of these words are related in meaning and others were not and the general finding was that it was faster to make lexical decision when the words were related than when they were not.

In the current experiment this finding is extended in several ways. First the letter strings are not presented simultaneously. The first string follows the second either 300, 600 or 900ms later. This allows us to look at the growth and decline of spreading activation across time and has been used in many recent experiments to investigate the time course of spreading activation. Some theories also suggest that with time, semantic activation is actively inhibited to return the network to its ready state (Gernsbacher & Faust, 1991). Another modification is that in the original experiment the nonwords were simply fillers, but in this experiment the nonwords were created by altering one letter from either a related or unrelated pair of words. In this way we can look at the theoretically relevant issue of whether non-words can also, at least partially, activate the meaning of a similar real word (for a review of theories of word recognition see Whitney, 1998) This might explain the resilience of the human language comprehension system to input that is often garbled or incomplete. If so, we might also see differences in RTs for non-word related pairs in comparison to non-word unrelated pairs. The question is whether the related pairs will be faster, as in the word condition, or slower, if the activated word meaning makes it more difficult to reject the string as a non-word.

Return to top

Stimuli and Design  

The independent variables are 1) lexical status: word or non word, 2) semantic relationship: related, unrelated, and 3) delay: 300, 600, 900ms. There are 72 word pairs, 6 in each condition.

The stimuli were created by using the University of South Florida Free Association Norms by Douglas L. Nelson and Cathy L. McEvoy, & Thomas A. Schreiber which are available on the web at http://w3.usf.edu/FreeAssociation/ for the related pairs the associates with the highest ratings were used. Across conditions word class and word frequency across the conditions was controlled. The related nonwords were created by changing one phoneme (word sound) in the pair.  

Return to top

Task

The participant is presented with 72 pairs of words in a difference random order. The task is to decide as quickly and accurately as possible whether both of the letter strings are real words.

Return to top

Analysis

The mean reaction times for all subjects can be found at the bottom of the Summary sheet in the excel spreadsheet. The experiment is  2 X 2 X 3 within subject design and the full design can be analyzed with a repeated measures ANOVA. Depending on interest, the data from the word condition can also be analyzed separately in which case one would expect a main effect of semantic relationship (related faster than unrelated). If spreading activation grows with time then one would also expect a main effect of delay condition. This effect could be followed up with posthoc tests. In the full design, higher order interactions would be expected. For example, if the words benefit more than nonwords from the semantic relationship then this would result in a two way interaction.

Data Format in Downloadable csv file
The data obtained via the Download data link on the PsychExperiments homepage are comma delimited.  The first entry (up to the first comma) is the affiliation of the research participant.  Typically this is the class to which the participant belongs.  Non-affiliated participants use the default entry of Interested Person. Next is the randomly assigned ID code for the research participant. The third and fourth entries are the data and time (24-hour format) on the server when the data were received.  Next come the raw data for the experiment organized by trial number. There is a trial number (1 to 72), a text entry for the lexical status (Word or Non-word), an integer entry for accuracy (1=correct, 0=incorrect), an entry for semantic relationship (related, unrelated), an entry for the number of seconds taken to make a decision on the trial (to 3 decimal places), and last a decimal value for the delay time.  Each of these entries is separated by a comma.  The table below shows the format trial data in tabular form.  Not included are the user information (affiliation, ID, data, and time) which precede these data as shown in the text line below the table. 

Return to top

References  

Gernsbacher, M. A., & Faust, M. E. (1991) The mechanism of suppression: A component of general comprehension skill. Journal of Experimental Psychology: Learning Memory and Cognition, 17, 245-262.

Meyer, D. E., & Schvaneveldt, R. W. (1971). Facilitation in recognizing pairs of words: Evidence of a dependence upon retrieval operations. Journal of Experimental Psychology, 90, 227-234.

Whitney, P. (1998) The Psychology of Language. Houghton Mifflin: NY