Reaction to Wood & Cowan (1995) by Richard Thripp
EXP 6506 Section 0002: Fall 2015 – UCF, Dr. Joseph Schmidt
September 15, 2015 [Week 4]
Wood and Cowan (1995) indicate that they are following up and improving on an old research study that was “conducted rather casually” with a tiny sample size (p. 255). Wood and Cowan’s participants listened to two channels of unrelated, monosyllabic words with stereo headphones, and were asked to attend to and repeat (“shadow”) only the female voice, while being instructed to ignore the male voice in the left earpiece. Unbeknownst to them, the male voice would say their name at either the 4 or 5 minute mark, and the name of another “yoked control participant” at either the 4 or 5 minute mark (p. 256–57). In the experimental condition, 9 of 26 participants noticed their name in the irrelevant channel, and 5 of these 9 participants made a mistake in repeating one or more of the two words before or three words after, compared to a much smaller proportion of errors among the other participants (p. 258). Interestingly, the 9 who noticed had a much higher mean response lag on the second word after their name—approximately 950 ms as compared to 675 ms in the next highest category, possibly indicating distraction (p. 259).
While this may be a compact and nicely structured study, the generalizability is limited—it is not similar to the “cocktail party” analogy at all. All words used were monosyllabic, and participants were specifically selected who had monosyllabic names—a highly unrealistic scenario, given the plethora of common disyllabic names. The attended channel was always in a female voice and the irrelevant channel in a male voice—given that higher pitched voices may be easier to hear, it would have been interesting to see the authors switch this up. Both channels played words simultaneously and at a rate of exactly one word per second (p. 257), which is not generalizable to a cocktail party, nor even most human conversation. I was surprised that while the authors were careful to play half of participants’ names at the 4 minute mark and others at the 5 minute mark, they did not try switching the ears (the attended channel was always the right earpiece). Furthermore, the entire experiment was only 5 ½ minutes—placing the stimulus so late in playback could produce different results from placing it toward the middle or beginning, although a Fisher’s exact test indicated no difference between the 4 and 5 minute conditions.
I cannot understand why 5 participants were rejected due to not having yoked control participants (p. 256). Why not just select monosyllabic names at random from a list of common names? In fact, I am not sure of the necessity of having yoked participants at all—selecting names randomly could have worked for all participants, freeing the authors up to manipulate some other variable. The authors’ admit that “the order of words was otherwise [besides the insertion of two names] identical across participants” (p. 257), but this could have been varied by experimental condition. There is also a gender bias that is not addressed: 25 (73.5%) of participants are male and only 9 (26.5%) are female. The authors could have selected equal numbers per gender, and could have broken out the results by gender.
I would like to have seen more discussion regarding the fact that none of the 26 experimental participants recalled hearing the yoked control participant’s name. This may be an indicator that monosyllabic names and words are not differentiated by our brains like our name is, but this possibility was not explored. It would even be interesting to conduct an experiment where the irrelevant channel consisted primarily or completely of monosyllabic names, to see whether this is noticed and whether a similar proportion of participants notice their names. Notice that 85% of participants were not even able to recall a specific word from the irrelevant channel, and 62% did not volunteer that it was in a male voice, even though all were asked for information about the channel’s content (p. 257). Were some participants just listening better than others, or not following the directions precisely? Participants who noticed their names made fewer errors on average: 17.0 versus 20.5 (p. 257). While this difference was not significant, recall that participants who noticed their names made more errors in the three words immediately after their names (p. 258–59). Perhaps if the errors immediately after their names were partialed out, a significant difference would have been found? We may never know.
There are many other conditions the researchers could have tried. While a sample of 34 is generally sufficient for a cognitive experiment, this was a very short and simple experiment that required little time or energy from participants. It would be nice to see the authors use a larger sample size and try more interesting experimental conditions, rather than rejecting 6 participants (p. 256) due to a shortage of names (n = 5) and due to an experimenter mistakenly letting the cat out of the bag (n = 1), albeit the latter is my speculation.
Reference
Wood, N., & Cowan, N. (1995). The cocktail party phenomenon revisited: How frequent are attention shifts to one’s name in an irrelevant auditory channel? Journal of Experimental Psychology, 21(1), 255–260.