Sunday, July 17, 2011

The Google Stroop Effect?

Notice the logo is multi-colored (as pointed out by Neurobonkers). Seeing "Google" printed in a solid color (or in any other font, for that matter) would likely result in a Stroop effect, or a slower response time in identifying the color of the font, relative to that of a neutral word.

Is Google making us stupid?

That question, and its original exposition in The Atlantic, has been furthering the career of Nicholas G. Carr. His subsequent book, The Shallows: What the Internet Is Doing to Our Brains, expanded upon his broader thesis that the internet is damaging to our cognitive capacity and the way we think. Numerous writers, both pro and con, have debated whether the internet and social networking sites (and computers in general) are harmful, so I won't belabor that point here. Instead, I'll cover a new article in Science that purportedly found Google Effects on Memory (Sparrow et al., 2011).

Cognitive Consequences of Having Information at Our Fingertips

The paper by Sparrow et al. (2011) conducted four experiments to determine whether the ability to access previously learned information reduces the effort put forth in remembering and retrieving the information. Specifically, the authors view the internet as a form of transactive memory, a means to offload some of the daily cognitive burden from our brains to an external source. Or, as succinctly expressed in ars technica, why bother to remember when you can just use Google?

This is nothing new, nor is it something dependent on the internet. In 1985 Wegner et al. (PDF) examined the way that married couples can have a division of labor along the lines of which facts to remember (Bohannon, 2011):
For example, a husband might rely on his wife to remember significant dates, while she relies on him to remember the names of distant friends and family—and this frees both from duplicating the memories in their own brains. Sparrow wondered if the Internet is filling this role for everyone, representing an enormous collective act of transactive memory.
Another example is illustrated by the phenomenon of the open book test. If students know they can use their textbooks to answer questions on an exam, they may put forth less effort into rote memorization of facts, and may instead learn the organization of each chapter, familiarizing themselves with where particular facts are located within the text. That indeed is what was demonstrated in Experiments 3 and 4, but in terms of accessing the information online or from a computer's hard drive.

The Google Stroop Effect

Experiment 1 asked whether the participants were primed to access computer-related words when faced with difficult trivia questions, relative to when they answered easy trivia questions (examples below).
Appendix A: Easy Questions

1. Are dinosaurs extinct?
2. Was Moby Dick written by Herman Melville?
3. Is the formula for water H20?
4. Is a stop sign red in color?
5. Are there 24 hours in a day?
. . .
16. Does a triangle have 3 sides?

Appendix B: Hard Questions

1. Does Denmark contain more square miles than Costa Rica?
2. Did Benjamin Franklin give piano lessons?
3. Does an Italian deck of card contain jacks?
4. Did Alfred Hitchcock eat meat?
5. Are more babies conceived in February than in any other month?
. . .
16. Is a quince a fruit?
The way the authors assessed automatic priming of internet- and computer-related words is by using a modified version of the ever-popular Stroop test. Name the font color of these words but don't read the words themselves:
  • RED
  • BLUE
Now do the same for this set of words:
  • RED
  • BLUE
Bet you were faster for the first set. That's because reading is a much more automatic process than naming the ink color in which the words are printed. This conflict between response options produces interference and slows reaction times (RTs) in the task.

The modified Stroop task used by Sparrow et al. relied on attentional salience rather than response conflict. Instead of color words, the participants viewed words related to computers and search engines, or words not related to these things:
This color naming contained 8 target words related to computers and search engines (e.g., Google, Yahoo, screen, browser, modem, keys, internet, computer), and 16 unrelated words (e.g., Target, Nike, Coca Cola, Yoplait, table, telephone, book, hammer, nails, chair, piano, pencil, paper, eraser, laser, television).
First off, you'll note that there are twice as many control words as there are computer words1. More importantly, you'll also notice that the unrelated words included prominent brand names (some of which are strongly associated with a particular color) and a grab bag of nouns from different semantic categories (furniture, tools, writing implements, musical instrument, etc.). The Google logo is multi-colored (as we've said before), and the current Yahoo logo is purple (it used to be red).

Hmm. So already we're looking at quite a confound. Nonetheless, the authors expected a larger Stroop effect for the search engines for different reasons:
In this case, we expect participants to have computer terms in mind, because they desire access to the information which would allow them to answer difficult questions. Participants are presented with words in either blue or red, and were asked to press a key corresponding with the correct color. At the same time, they were to hold a 6 digit number in memory, creating cognitive load.
Why? Why oh why did the authors want to create a cognitive load during the Stroop? This turns the whole study into a dual task experiment, requiring the participants to multi-task: a key press for red or blue (which requires retrieval of stimulus-response mappings) while remembering a 6 digit number. A rationale was not given for this particular manipulation. In addition, the more classic Stroop task measures voice onset times, or the RTs to verbally name the color. In that case, the participants don't have to remember which key corresponds to a red or blue response. However, this latter issue is much less objectionable than the choice of brand names.

Target is strongly associated with the color red, as is Coca Cola. On the other hand, Nike is most often seen in black, but can also be found in other colors.

Thus, we have color-consistent Target and color-inconsistent Target as experimental stimuli while Google and Yahoo are always color inconsistent (unless you remember the red Yahoo logo from before May 2009). The branded experimental stimulus displays might have looked something like this (but with the words presented one at a time):

If attention is drawn to the search engine words to a greater extent in the difficult trivia condition, this would be manifest as slower RTs compared to the other brands (i.e., the Google Stroop effect). And that's more or less what was observed (with important caveats).

Fig 1 (Sparrow et al., 2011). Accessibility of brand names (as measured by color-naming reaction time) following blocks of easy or hard test items. Error bars are ±SEM.

RTs were substantially slower for Google/Yahoo when the questions were hard than when they were easy, suggesting the terms were more accessible in the former condition. As well, Google/Yahoo RTs were slower than Nike/Target in the hard condition (p<.003), BUT this was also true after the easy condition (p<.005). This could be because of the ridiculous color confound, but the authors state:
Although the concept of knowledge in general seems to prime thoughts of computers, even when answers are known; not knowing the answer to general knowledge questions primes the need to search for the answer, and subsequently computer interference is particularly acute.
This interpretation dilutes their hypothesis, because who needs to access the internet to know there are 24 hours in a day? At any rate, although the Google Stroop Effect did not provide strong evidence for Google's specific effects on memory, the other experiments demonstrated that the concept of transactive memory can be extended beyond families and co-workers to include the internet and computer hard drives. Is this a bad thing? No, said Sparrow and colleagues, who are realistic about the desire to remain online:
It may be no more that nostalgia at this point, however, to wish we were less dependent on our gadgets. We have become dependent on them to the same degree we are dependent on all the knowledge we gain from our friends and coworkers—and lose if they are out of touch. The experience of losing our Internet connection becomes more and more like losing a friend. We must remain plugged in to know what Google knows.


1 This would generally work against the Stroop effect, because the subjects would more easily habituate in the condition with fewer words.


Bohannon J (2011). Searching for the Google Effect on People's Memory. Science 333:277.

Sparrow, B., Liu, J., & Wegner, D. (2011). Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips Science DOI: 10.1126/science.1207745

Wegner, D. M., Giuliano, T., & Hertel, P. (1985). Cognitive interdependence in close relationships. In W. J. Ickes (Ed.), Compatible and incompatible relationships (pp. 253-276). New York: Springer-Verlag. PDF

Subscribe to Post Comments [Atom]


At July 17, 2011 4:10 PM, Anonymous Anonymous said...

I suppose you could test for the confound by comparing the computer associated brands with the other computer associated words. Or you could simply remove the brand words from the analysis.

Not entirely related, but I'd expect heavily advertised brands to be an interference nightmare; I know I personally can't read the word capital-T "Target" without struggling to resist their eerie subconscious influence - I'd like to see a test that measures such interference effects in branding words. Might be worth looking at the raw data to see if there's a difference between the branded and unbranded words.

The cognitive load was likely an attempt to enhance any possible effects by crowding short term memory; the effect might have been slight enough on cognitive load that it would not have been visible with full cognitive resources. Unlike the normal stroop, we're not as likely to trip our own processing in performing a color-task.

At July 17, 2011 5:45 PM, Anonymous Emmy said...

Interesting and, in my opiniion, distrubing as well. Frontline's program called Digital Nation featured a Harvard professor who said she could no longer assign books over 200 pages long (this is at HARVARD) due to low attention spans. While analysts of computer-age effects may find these developments charming, I wonder how we're going to make Galileo or Einstein-like science discoveries in an age where we're slaves to our computers.

At July 18, 2011 2:26 PM, Blogger The Neurocritic said...

hypotheticallycorrect - In the Supporting Online Material, the authors did an analysis that omitted the brand name words:

"Taking out the 4 terms (Google/Yahoo and Target/Nike) which yielded an interaction with easy/hard questions (F(1,66) 5.52, p <.03), the interaction between computer and general terms and easy/hard questions remains significant F(1,66) 9.49, p <.004."

The SD for the Google/Yahoo RTs was very high:

" engine brands after both easy (M = 638 ms, SD = 260 ms) and hard questions (M = 818 ms, SD = 517 ms) created more interference than general brands after easy (M = 584 ms, SD = 220 ms) and hard (M = 614 ms, SD = 226 ms)."

The cognitive load manipulation lacked justification, in my view. The one paper they cited on the "modified Stroop" used emotional words and no working memory load. So basically, Google/Yahoo and computer-related words are more accessible than unrelated words after answering trivia questions if your working memory is taxed.

Emmy - I think this particular study has more to say about flexibility in how we remember things, than about low attention span.

At August 30, 2011 2:58 AM, Blogger bob said...

I have always maintained that each individual needs to be encouraged to become an independent researcher - theorist if the world is goping to advance and yes Einstein would be very saddened by todays statistics within academia.


Post a Comment

<< Home

eXTReMe Tracker