Skip to main content
Category: Effects
Type: A visible empty slot in something you did not need pulls a further purchase that fills the slot. The counted pull is a hypothetical beer choice in 2017
Origin: A Chinese newspaper column, Life Times (Shengming Shibao), 16 June 2015, tells the cage story. The counted choices are Kate Barasz, Leslie K. John, Elizabeth A. Keenan, and Michael I. Norton, Journal of Experimental Psychology: General 146, no. 10 (October 2017), pages 1460–1477
Also known as: Birdcage logic, the name in that same popular telling. The alias is the story, not a second experiment
Quick Answer — The Birdcage Effect is the pressure a visible empty slot puts on the next purchase. A 2015 Chinese column tells of an empty cage, guests who ask where the bird is, and a man who buys a bird to stop explaining. That column prints no count. In a 2017 picture test, people who saw beers sitting in a four-pack carrier most often ended at four beers in all. People who saw the same bottles loose, and who were reminded that beer often comes in sixes, most often ended at six.

​
What is the Birdcage Effect?

The Birdcage Effect is the pull of a visible hole in a possession you did not ask for: the hole makes one more purchase feel like the way to finish the object.
The modal number of beers purchased among pseudo-set participants was four—the number needed to complete the pseudo-set.
That sentence is from Study 5 of Barasz, John, Keenan, and Norton, published in October 2017. A mode is the most common total, not the share of people who chose it. The authors call an arbitrary grouping a pseudo-set. In that study the grouping was a four-bottle carrier with empty slots. Diderot Effect is a different spiral: one new object makes the older objects around it look too plain, so those neighbors get replaced. Zeigarnik Effect is about memory for an interrupted task. Neither one is an empty slot in a carrier.

​
Birdcage Effect in 3 Depths

  • Beginner: An empty cage on a desk invites the question, “Where is the bird?” The 2015 column ends with buying a bird. It also says you can give the cage away. It prints no share for either ending.
  • Practitioner: When a holder, a starter kit, or a carrier shows an empty place, write the price of the missing piece before you buy it. In the 2017 pictures, the usual six and the carrier’s four were two different modes.
  • Advanced: The mode of four is the most common total in the carrier condition. It is not a printed percentage, and it is not a receipt. The completeness scores of 39.7 and 64.1 come from a separate picture of five beers, not from this choice.

​
Origin

The story read here is the web reprint of a column in Life Times, 16 June 2015, by Ju Jinmin, under the heading “Birdcage effect: jump out of habitual thinking,” in the section “Psychology clinic.” The reprint says that in 1907 a Harvard psychologist named James bet his friend Carlson that he could get Carlson to keep a bird. A few days later James sent an elegant birdcage. Carlson answered that James should not bother, because Carlson still would not keep a bird. After that, guests who visited asked, “Professor, when did the bird you kept die?” Carlson explained again and again that he had never kept a bird. The guests looked as if they did not believe him. To stop the endless explaining, he bought a bird. The column then defines the pattern in one sentence: after you happen to receive something you did not need, you cannot help adding more things you do not need. It also gives a second exit. Not every pretty cage should hold a bird. Giving the cage to someone else is possible too. The column goes on to careers and marriages. That next paragraph is the columnist’s advice. It is not a count of job changes or of cages given away. The column does not call Carlson a physicist, does not say either man had retired, and does not say the cage was a birthday gift. It does not cite a book or a paper by James. William James’s own 1907 papers, read here, are a different object. The Harvard Crimson of 23 January 1907 reports that on the day before, his last class, Philosophy D, gave Professor W. James a silver loving cup. The inscription dates the gift 22 January 1907 and calls the class his last in philosophy at Harvard. The note says he would live in Cambridge and might still lecture sometimes, and that the Harvard Corporation had not yet decided what post he would hold. In a letter from Cambridge dated 21 February 1907, to Moorfield Storey, James writes that his retirement does not take place until the end of the year, and that the papers have “railroaded me out too soon.” Neither the Crimson note nor this letter names a bet, a birdcage, or a friend called Carlson.

​
Key Points

Keep the column, the four-pack choice, and the separate completeness picture in three piles. A guest’s question is not a mode. A mode is not a percentage.
1

Read the column as a story with two exits and no rate

The 2015 reprint ends the anecdote with a purchase: Carlson buys a bird so he can stop explaining. The same column then says a pretty cage need not hold a bird, and that giving the cage away is possible. It prints no number of guests, no share who bought a bird, and no share who gave a cage away. Throwing the cage out is not the sentence it prints. Sunk Cost Fallacy is staying because money or effort is already gone. The cage in the column was a gift. The column does not name a price.
2

Keep 22 January 1907 apart from the bet

The dated Harvard records read here are William James’s last class and his February letter about when retirement would actually start. They support the year 1907 and the name of a Harvard psychologist. They do not support a wager, an empty cage, or Carlson. A later retelling that has James already retired, and already winning the bet, is ahead of the February letter.
3

Use the carrier, not the six-pack reminder, as the thing that moved

Study 5 told everyone that beer is often sold in multiples of six, that each bottle is priced alone, and that there is no discount for buying more. People imagined one, two, three, or four bottles already in the cart. One group saw those bottles loose. The other saw them inside a four-pack carrier, so the empty slots showed. The most common total was six among the loose bottles and four among the carrier pictures. A Kolmogorov–Smirnov test on those total distributions gave Z = 3.15, p below .001. A second test, on how many more bottles people added, gave Z = 1.47, p = .027. At each starting count, the most popular add-on was the one that reached six in the loose group and four in the carrier group. The figure’s bar heights are not printed as percentages on this page. Unit Bias is finishing the portion that is already in front of you. These people were choosing how many more bottles to add. They were not eating a serving.
4

Keep the five-beer picture off the four-pack choice

A separate Mechanical Turk sample, N = 273, mean age 37.3 (SD 11.3), rated five bottles and bought nothing. Five loose bottles scored 64.1 for completeness (SD 29.8). The same five, sitting in a six-pack carrier, scored 39.7 (SD 31.3). The comparison is t(271) = 6.60, p below .001, d = 0.80. The scale ran from 0 to 100, with the numbers hidden from the rater. Five of six is not one empty slot in a four-pack, and 39.7 is not a share of shoppers. Endowment Effect is a higher price for a thing once you own it. This slider is a judgment that the picture looks unfinished.

​
Applications

Name the empty slot, the missing piece, and which exit you are actually choosing. Filling, giving the object away, and doing nothing are three different acts.

Price the missing piece before you fill the slot

A cage, a frame with one empty opening, or a starter kit with a blank compartment is the column’s object. Write what the missing piece costs, and whether you wanted it before the holder arrived. The 2015 column’s purchase is a bird bought to stop questions. It is not a receipt you can copy.

Ask which total the picture is suggesting

In the 2017 wording, six was the stated custom and four was the size of the carrier. The mode followed the picture people saw. Before you match a carrier, a tray, or a “set of four,” write the number you wanted when the slots were not on the screen. The study was a picture. It was not a checkout.

Treat giving the object away as its own exit

The column’s second exit is to give the cage to someone else, so the empty slot leaves the room. Study 5 allowed people to add zero further bottles. It did not offer a button that removed the carrier. Adding nothing and giving the holder away answer different questions. Do not call either one the usual ending.

Mark inherited slots on a plan

A team board, a course page, or a budget sheet can show an empty box you did not choose. Write whether that box was a requirement or a picture of a set. The beer mode is not a hiring rate, a module-completion rate, or a line item. Escalation of Commitment is staying with a course of action that is already going badly. An empty box is not that history.

​
Case Study

In 2017, Kate Barasz, then at IESE Business School in Barcelona, and Leslie K. John, Elizabeth A. Keenan, and Michael I. Norton, at Harvard Business School, published Study 5 in the October issue of Journal of Experimental Psychology: General. The paper’s last page says the revision was received on 7 April 2017 and accepted on 8 May 2017. The question was whether an arbitrary carrier would change how many bottles people said they would add, even after everyone had been told that beer is often sold in sixes. Eight hundred and two adults on Amazon Mechanical Turk, mean age 35.6 (SD 11.8), read a grocery scene. The supermarket sold bottles one by one. Each bottle had its own price. There was no discount for volume. The store offered optional carriers. People were told they had already placed one, two, three, or four bottles in the cart, and they saw a picture. Half saw loose bottles. Half saw the same counts inside a four-pack carrier. They then did two things, in either order: they rated how complete the picture looked, on a hidden 0-to-100 scale, and they chose how many more bottles to add. The menu ran from zero through six or more. The paper reports no effect of which question came first (p = .178 for the number added, p = .282 for completeness). The usual total and the carrier did not point at the same number. Among people who saw loose bottles, the most common total, bottles already shown plus bottles added, was six. Among people who saw the carrier, the most common total was four, the number that filled it. The two distributions of totals differed, Z = 3.15, p below .001. The authors write that the four-pack seemed to exert more influence than the six-pack reminder, even though the carrier group had heard that reminder too. They also report a regression in which the number already in the cart, the framing, and the interaction of those two predicted how many more bottles people wanted. The interaction coefficient was −0.21 (SE 0.10, p = .045). On the completeness slider, the framing coefficient was −27.95 (SE 5.29, p below .001), and the interaction was 8.60 (SE 1.93, p below .001). The paper’s sentence is that, for the same number of bottles, the carrier looked less complete, except once four bottles filled it, when the pattern reversed. Completeness, entered into the same regression, carried a coefficient of −0.02 (SE 0.002, p below .001), and the framing-by-count interaction fell to −0.01 (SE 0.10, p = .892). A bootstrap of 10,000 samples put the indirect effect in a 95 percent interval from −0.29 to −0.11. The paper calls the mediator’s effect size −0.36 in the carrier condition and −0.16 in the loose-bottle condition. Those are coefficients, not bottles and not shoppers. The lesson for the Birdcage Effect is narrow. A pictured hole, with no guest in the room and no discount on the label, was enough to move the most common total from the stated custom of six to the carrier’s four. The lesson stops at that hypothetical menu. The authors mention Whole Foods as an example of a retailer that lets shoppers put loose beers into a carrier. The 802 people were not that shop’s customers, and no price or receipt is in the study. Adding zero was allowed. Removing the carrier was not. A bird, a guest’s question, and a cage given away are the 2015 column. They are not rows in this table.

​
Boundaries and Failure Modes

The pull needs a hole that still looks fillable. Barasz and colleagues write that if finishing the set is too hard, or simply impossible, this kind of framing may lose its pull. They chose tasks a person could finish with a moderate effort. A carrier you cannot fill, or a missing piece you have already decided you do not need, is outside the pattern they tested. They do not print a cutoff for “too hard.” A full holder is a different picture. In Study 5 the completeness gap reversed once four bottles occupied the four-pack. The empty slot, not the brand and not the mere presence of a carrier, is what the paper ties to the unfinished look. A cage that already has a bird is not the stimulus in the column either. The common misuse is to export the story or the mode. The column prints no probability that a person buys the bird, throws the cage away, or gives it away. The mode of four is the most common total in one hypothetical condition. It is not the percentage who chose four, because those bar heights are not printed here. It is not a share of supermarket trips, households, or teams. The column’s later lines about changing jobs or leaving a marriage are the columnist’s advice. They are not a result of Study 5.

​
Common Misconceptions

The year 1907, the purchase of a bird, and the beer totals get stacked into one fact. They are three different texts.
The 2015 column names a Harvard psychologist called James and a friend called Carlson. It cites no book and no experiment. The Harvard records read here show William James’s last class on 22 January 1907 and, on 21 February, his statement that retirement was still ahead, at the end of the year. Those records do not contain the bet, the cage, or Carlson.
The column says Carlson bought a bird, and that people in this spot cannot help adding more unneeded things. It prints no share and no sample. Study 5 prints a mode, four against six, for imagined beers. A mode can be the most common total without being a majority you are allowed to quote, because the percentage is not in the prose. Nobody in that study was offered a bird.
The Diderot pattern starts from a finished new object that makes older neighbors look shabby, and the spending replaces those neighbors. The cage pattern starts from a hole in the new object. The 2017 result moves a total toward the size of the carrier. It does not show people replacing the rest of a room. Giving the cage away, the column’s second exit, is not a cascade of upgrades either.
These pages sit next to the empty slot. Each one measures a different act.

Diderot Effect

One new possession makes the older things around it feel mismatched, and further purchases replace those neighbors. The cage is the hole, not the scarlet gown.

Zeigarnik Effect

Unfinished tasks can linger in memory. Barasz and colleagues cite that line of work for the feeling of incompleteness. This page does not move a recall score onto the beer totals.

Unit Bias

The unit in front of you can set how much is eaten. The four-pack here set how many more bottles people said they would buy. It was not a serving.

Sunk Cost Fallacy

A past spend is hard to walk away from. The column’s cage was a gift, and Study 5 printed no price and no discount. Filling a hole is not the same claim as refusing to quit.

Endowment Effect

Ownership can raise the price you ask. The 0-to-100 slider asked whether the picture looked finished. It did not ask for a selling price.

​
One-Line Takeaway

When an empty slot shows, write the missing piece and its price, then choose fill, give the holder away, or leave the hole. The 2017 mode tells you a carrier can move the most common total. It does not tell you which exit you owe.