stuff
blog
about me
writings
old journal
applesauce obsession
harem quotes
n00n quotes
linky-dinks
email

* photo is gasworks park, seattle, taken by kevin.

12/10/2002
Acoustic Similarity, Letter String Length, and Their Effects on Ease of Recall
Emily N. Cannon
With Andrew Freeling, Amy Klehm, Leah Lee, and Chisato Maeda
University of Washington

The capacity of short-term memory (STM) is not as simple as Miller's 7 +/- 2 paradigm (1956) and may be affected by the type of stimuli. In our study, we looked at the way acoustically-similar phonemes interacted with STM capacity using string length and a letter recall task. Forty-five students were read one set of acoustically-similar letter strings and one of acoustically-different strings, each containing three four-letter strings and three six-letter. Results for the interaction of sound and string length were t(44) = -6.25*, p<.05. Recall was lowest for acoustically-similar six-letter strings, evidence that STM capacity is decreased by acoustic similarity.

Some things are harder to remember than others. How do people remember bits of information? Information presented in the environment, such as words, must go through short-term memory to be processed and recalled, even immediately after the fact. The problem is that short-term memory has a limited capacity. This has been tested for many years using, among other things, digitspan tasks in which subjects were read and asked to recall longer and longer strings of numbers. It is from this body of empirical research that we believe short-term memory capacity to be about seven digits, plus or minus two (Miller, 1956). Numbers, however, may not be processed the same way letters are, nor are there as many differences in the way they sound. Our research was concerned with the way things that sound alike may be more difficult to recall than those that do not sound alike.

Though a simple idea, this problem could have implications for many areas, especially language. The ability to recall letters themselves in sequence may affect development of language, as in how certain letter combinations become common where others do not. It may also affect spelling ability and the ability of non-native speakers to learn a second language. On the other hand, similar-sounding letters could make it easier to recall, the way rhymed poetry seems more readily recalled than blank verse. A good illustration of why we did not believe this to be the case in our research is a tongue twister like Peter Piper. The rhyme is catchy, but hard to remember and reiterate correctly because the acoustically-similar words are confusing.

Previous research by Peatman and Locke (1934) examined the methodology of the popular digitspan test. They were concerned with the reliability of the tasks and which method is most effective. Variables included number of series, oral versus visual presentation of stimulus, regular versus irregular speed of presentation, oral or written reporting by subjects, and basal or weighted scoring. According to their research, the most reliable method was oral presentation of at least four series, and the other variables did not differ that much except that weighted scoring (a percentage of each correct score obtained after the basal score) positively favored the subjects.

There are visual and acoustic components to short-term memory (Laughery, Welte, and Spector, 1973). Researchers hypothesized that different types of errors (visual or acoustic) are likely to occur in different positions in a 10-digit letter recall task varying with presentation rates (.3, 1, and 2 seconds). More visual and acoustic errors occurred in the last digits of the string, acoustic errors occurred more frequently than visual, and faster presentation interacted with acoustic and visual errors in the first part of the string.

Our research was specifically focused on comparing the accuracy of recall for similar sounds versus different sounds with the amount of information being presented. We hypothesize that it will be more difficult to recall similar sounds accurately than different sounds; additionally, it will be more difficult to recall larger amounts of information. We decided to test letters for their simplicity, since vocabulary and word usage can vary among even native English-speakers. Letter-string length is something that has already been used extensively (Laughery, Welte, and Spector, 1973), and we believed there would be an interaction between the two variables.

Our methods are fairly simple. We will read six similar-sounding letter strings and six different-sounding letter strings to a group. Within each of those sets, three strings will be four letters long and three will be six letters long. Half the subjects will hear the different-sounding set first, while half will hear the similar-sounding set first. They will write their responses on paper provided for them, which we then score. Due to time constraints, we are unable to test as many independent variables and levels thereof as we would like, but we determined this simplified method would provide the strongest results with the least amount of complication. We predict that similar-sounding letter strings will be recalled less accurately than different sounding letter strings, the six-letter string will have less accurate recall than the four-letter string, and that there will be an interaction between the two resulting in similar-sounding 6-letter strings being the least accurately recalled.

Participants
Forty-five University of Washington students participated in this study. All were enrolled in Psychology 331, though about one third in an 8:30 class, one third in a 10:30 class, and one third in a 12:30 class. The task was administered during normal class time.

Material
Letter strings were read aloud (Figure 1) by a live experimenter. The letters were chosen for their acoustic similarity (B, C, D, E, G, P, T, V, Z) or from a selected set of acoustically-different letters (F, H, L, O, Q, R, K, W, X) and arranged in strings to avoid forming word patterns or locating the same letter at the end of one string and the beginning of the following string. The string lengths of four and six were chosen after pilot studies indicated that seven or higher resulted in too many errors to provide useful data. Participants were given a printed response sheet and asked to provide a writing tool.

Design
The study is a within-subjects 2 (sound) x 2 (string length) design. For each variable there were two conditions. In the first variable, there were similar-sounding letter strings and different-sounding letter strings. In the second variable, there were four-letter strings and six-letter strings.

Procedures
Each participant performed in 12 trials, six for each category of letter string (similar- and different-sounding). Within each category, three strings were four letters long and three were six letters long. Participants were read a list of letters and asked to write down as many as they could recall. Half the participants heard the similar-sounding list first and half heard the different-sounding list first. The same speaker administered all trials, and her verbal instructions read:

"I will read a list of letters. Your task is to write all the letters in the list, in the order given. I will read one list at a time, starting with Number 1. Please be sure to fill in every blank space on your sheet. Do not begin writing until I have finished reading each list. Each list will be read only once."

Recall we are interested in the difficulty of recalling similar-sounding items and larger amounts of information. We tested this by asking subjects to recall four- and six-letter strings of acoustically similar and different letters. Subjects did have more difficulty recalling similar-sounding items and longer strings and had the worst recall for similar-sounding six-letter strings.

For similar vs. different sounding phonemes, the mean number correct for similar was 3.22, while for different it was 5.33, with results from the t-test of t(44) = 10.89*, p < 0.05.

For 4-letter vs. 6-letter strings, the mean number correct for four letters was 5.47, while for six letters it was 3.09, with a t-test result of t(44) = 14.59*, p < 0.05.

For the interaction effect (Figure 2), the mean difference in letter string sounds was 0.44, in letter string length, 1.67, and a t-test result of t(44) = -6.25*, p < 0.05.

Our hypotheses, that the accuracy rate for acoustically similar phonemes and six-letter strings would be lower, and that there would be an interaction between the two variables, were supported. Similar-sounding letter strings were recalled less accurately than different-sounding letter strings; six-letter strings were less accurately recalled than four-letter strings; an interaction of both main effects resulted in the six-letter similar-sounding letter strings being the least accurately recalled of all. We are able to generalize these findings to the population because our sample was large enough to assume normality. This suggests that short-term memory is affected by factors like auditory processing.

In a larger context, these results could mean a number of things. For one, it is evidence of auditory confusion. This could also have implications for language development, learning a second language, spelling difficulties, and learning disorders like dyslexia.

Our results relate to studies that found acoustic errors occurred more frequently than visual (Laughery, Welte, and Spector, 1973). That study looked at both visual and acoustic stimuli, but since our study relied on acoustic stimuli alone, we found errors in shorter strings than the previous study, which had a 10-letter span. We were more concerned with the acoustic component of short-term memory.

We were limited by our set size for time constraints and the sake of not fatiguing our subjects. We found in pilot studies that subjects were fatigued after the original experimental design, which consisted of three levels of the string length variable and had strings of five, six, and seven letters. Subjects may also have been tired of this sort of task because our experiment was one in a sea of similar recall tasks administered on the same day. There is also the matter of presentation: stimuli was presented by a live human, so it is possible there was variability in the speed and volume of the reading which may have affected performance.

In the future, we would analyze a third and possibly significant independent variable, order. While half the subjects were read the different list first and the other half heard the similar list first to rule out order effects as a confound, the possible order effects could have been interesting to explore. Pilot data indicated that similar-sounding strings were easier to recall when read first. In future studies, researchers may want to consider order effects.

Additionally, we wonder about testing other types of similar-sounding letters, like s and c, which do not rhyme like our similar-sounding stimuli, but do sound similar. We would also want to explore the location and frequency of letters within each string and test more levels of the string length variable to better pinpoint this.

Other questions that may be of interest for future research include whether these results hold cross-culturally, particularly in languages with a radically different sound and structure from English, like Mandarin? What does it mean for people learning a new language, like an American learning Mandarin? What effects does this have on children learning language and spelling? How would subjects with a learning disorder like dyslexia perform on this or a similar task? Similar tests could be designed to explore some of these questions--a between-subjects design with some English-speaking subjects and some Mandarin-speaking, for example, or testing normal learners versus dyslexics.

Overall, we have evidence that auditory confusion affects short-term memory capacity in a substantial way. Uniqueness, the ability to stand out in a sea of sameness, bodes well for someone trying to be noticed and remembered. The same goes for bits of information. While schoolchildren may trip over recitations of Peter Piper picked a peck of pickled peppers," they delight in the peculiarity of a poem like Lewis Carroll's famous "Jaberwocky," always remembering to beware that strange beast.

Laughery, K.R., Welte, J.W., & Spector, A. (1973). Acoustic and visual coding in primary and secondary memory. Journal of Experimental Psychology, 99(3), 323-329.
Miller, G.A. (1956). The magical number seven plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63, 81-97.
Peatman, J.G., & Locke, N.M. (1934). Studies in the methodology of the digit-span test. Archives of Psychology, 25(167), 5-31.