Google Effect

​

What is Google Effect?

Google Effect is lower later recall of information people expect to reach again, and higher recall of where that information was kept when the place test gives a cue the fact test does not.

The results of four studies suggest that when faced with difficult questions, people are primed to think about computers and that when people expect to have future access to information, they have lower rates of recall of the information itself and enhanced recall instead for where to access it.

That sentence is the 2011 abstract. It does not mention a cue. The cue is in the body of the fourth study. People read trivia, typed it, and were told each entry had been saved in a folder with a generic name. They then had 10 minutes to write the statements. A later question named a word from the statement and asked for the folder. Mean recall of the folder was 0.49. Mean recall of the statement was 0.23. The authors write that the folder question had a cue the statements did not. Leaving a dose in a phone note has the shape of expecting access. It does not inherit 0.49 or 0.23. Those two proportions are mean shares of the statements in that fourth study. Spacing Effect is better later recall when two studies of the same item sit at least 1 second apart. A gap between two looks is not a belief that a file will still be there.

​

Google Effect in 3 Depths

• Beginner: When people were told the computer would erase what they typed, mean recall of 40 trivia statements was 0.31. When they were told it would save the typing, mean recall was 0.22. Both cells had no extra instruction to remember. The 0.31 and the 0.22 are proportions of statements, not proportions of people.
• Practitioner: If the later use is a blank page, the 2011 save message is the condition that recalled less. An added line that told people to remember did not have a significant main effect. The paper prints that conclusion and does not print a separate test statistic for the instruction alone.
• Advanced: The 0.49 folder score was cued recall. In the third study, recognition of the specific folder was 0.23, while recognition that a folder-saved statement was exactly the one they had read was 0.85. A color-naming delay after a hard question is a different contrast, and two later tests did not recover it.

​

Origin

The four studies, and the English title Google Effects on Memory, are the 2011 report. This page keeps the name Google Effect for that report. The plural title is the article, not a second finding. Betsy Sparrow was at Columbia University, 1190 Amsterdam Avenue, New York. Jenny Liu was at the University of Wisconsin–Madison, 1202 West Johnson Street, Madison. Daniel M. Wegner was at Harvard University, 33 Kirkland Street, Cambridge. The manuscript was received on 2 May 2011 and accepted on 27 June 2011. It was published online on 14 July 2011 and printed in Science 333 on 5 August 2011, pages 776–778. The author PDF read here prints the DOI 10.1126/science.1207745. Columbia University supported Sparrow. National Science Foundation grant BNS-0841746 supported Wegner. The paper places the studies next to transactive memory. It cites Daniel M. Wegner, in Theories of Group Behavior, edited by Brian Mullen and George R. Goethals (Springer-Verlag, New York, 1986), pages 185–208, and Wegner, Social Cognition 13, 319 (1995). Those two works were not re-read. No sentence from them is on this page. The 2011 paper’s own description is a mix of memory a person holds and memory the person can reach because someone else holds it. The paper then asks whether a search engine has become that outside store. The flag question in the introduction, whether any country has only one color in its flag, is the authors’ illustration. It is not reported as one of the trivia items.

​

Key Points

Four cuts keep a saved note from inheriting a proportion it did not measure. The color name, the erase message, the recognition question, and the cue each have their own number.

​

Applications

Use the contrast when someone treats any saved note, or any slow color name, as the proportion from one of the four studies. Keep 0.31, 0.22, 0.49, and 0.23 inside the questions that produced them.

​

Case Study

In the fourth study, Sparrow, Liu, and Wegner asked whether people would recall where a statement was kept more often than they would recall the statement. Everyone was led to expect that each typed trivia statement would be saved in a folder with a generic name, the same kind of name as FACTS in the third study. There was no practice trial. The paper says the names and the number of folders were never explicitly pointed out. People were not told that there were six names. They then had 10 minutes to write as many statements as they could. After that, an identifying word from a statement was given, and they had to type the folder name. For the ostrich sentence, the question was which folder the statement about the ostrich had been saved in. The dialog box was called Items. Folder names were not mentioned again after the original typing. Mean recall of the place was 0.49, standard deviation 0.26. Mean recall of the statement was 0.23, standard deviation 0.14. The paper prints this comparison as a between-subject t(31) = 6.70, P below 0.001, two-tailed. The two scores come from the same people. The authors also write the caveat in the same section. The folder question had a cue from the trivia statement. The statement recall did not. They write that they were not able to counterbalance the trivia and the folders so that the folders were as numerous as the statements, which would have been needed to balance the cued and uncued tasks. They then split each person’s statements into four kinds of outcome. Both the statement and the folder: mean 0.17, standard deviation 0.16. The statement but not the folder: 0.11, standard deviation 0.08. Neither: 0.38, standard deviation 0.24. The folder but not the statement: 0.30, standard deviation 0.16. The printed omnibus test is F(1, 31) = 11.57, P below 0.003. The paper says the comparison of the first two cells was not significant, and the comparison of the last two cells was not significant. The pattern they describe is that the folder score shows up when the statement does not. That description sits next to those two non-significant pairs. It is not a claim that 0.30 is reliably above 0.38. The lesson is a cued place score above an uncued fact score, among statements people expected to remain saved. It is not a rate at which a class remembers a chapter, and it is not the third study’s 0.23 for recognizing the specific folder when the folder names were on the screen.

​

Boundaries and Failure Modes

Google Effect is the wrong summary when a cued folder is treated as uncued place memory, when the third study’s low folder recognition is ignored, or when a color-naming delay is treated as a settled reflex. The first boundary is the cue. The authors write that the 0.49 looks unexpected because the statements were memorable and the folder names were not, and then they give the cue as the reason for caution. A file path remembered only after someone says “the ostrich one” is the cued question. A blank page is the uncued question. The standard deviation of 0.26 on the 0.49 is the spread of that cued mean. It is not a guarantee for each statement. The second boundary is the third study’s folder question. When the folder names were offered, specific-folder recognition was 0.23, standard deviation 0.14, against 0.85 for recognizing the exact wording of a folder-saved statement. The paper itself calls the folder recognition dismal next to the fourth study’s recall. Believing an entry was saved did raise memory for the fact that it could be reached, on the saved-or-erased question, with means 0.61 and 0.66 against 0.51 for erased. That is memory for the access tag. It is not memory for which of the six names was used. Von Restorff Effect is better memory for the one item that differs in kind from similar neighbors. Six generic folder names were chosen to be interchangeable. An odd neighbor is not one of those names. The third boundary is the color-naming contrast, and it is the common misuse. The supporting material, as quoted by Guido Hesselmann in PeerJ (2020), says 46 Harvard undergraduates, 28 women and 18 men, were tested. The main text prints t(68) and F(1, 66). Hesselmann writes that the printed t, F, and p values agree better with 69 people, and that the degrees of freedom are not right for either headcount. Colin F. Camerer and colleagues, in the supplementary information to Nature Human Behaviour 2 (2018), pages 637–644, write that the paper reports n = 46 but that n = 69 was actually used, and that the original authors said so only after the replication. That replication did not recover the color-naming result. The team did not receive the original materials. Wegner had died in 2013. Afterward, the original authors listed three design differences: a new six-digit number before each word in the original, against one number held for the whole block; 24 color-naming trials against 48; and mistakes kept in the original analysis, against mistakes removed. The supplementary information says the non-replication conclusion is unchanged for the first 24 trials and for an analysis that keeps mistakes. It also says the load difference cannot be ruled out. Planned power was 90 percent. Achieved power was 82.0 percent in the first collection and 80.7 percent in the pooled collection, because the sample-size error changed the estimated effect used for planning. Hesselmann, at the Psychologische Hochschule Berlin, then ran a closer procedure. Of 117 people tested, the reported test uses 89. The interaction contrast was t(88) = −1.04, p = 0.301. The Bayes factor BF01 was 5.07 in favor of no interaction. After easy questions, computer words averaged 901 milliseconds and general words 830. After hard questions, computer words averaged 821 and general words 825. The slowdown sat on the easy block. Those two replications tested the color name. They did not test the 0.31 against the 0.22.

​

Common Misconceptions

Three retellings turn a cued folder into uncued place memory, treat a remember instruction as a cure, or read any search as this report.

​

Related Concepts

The saved-note result sits next to a millisecond word decision, a study gap, a truth rating after repetition, memory for one odd neighbor, and the belief that a noticed thing happens a whole lot.

​

One-Line Takeaway