What the tunneling claim actually requires
Scarcity theory's tunneling hypothesis makes two linked claims about attention. Scarcity is said to increase focus on managing the scarce resource itself, which can produce positive outcomes, and this focus is said to come at the cost of attention to other potentially beneficial information, producing neglect. The second half is what makes the account distinctive: greater focus alone would not show a tunnel.
The influential evidence came from Shah and colleagues' 'Angry Blueberries' game, in which participants shot blueberries at targets and earned points for hits. Scarcity was manipulated by randomly giving participants large or small budgets of shots, making them 'rich' or 'poor'. The original authors inferred greater focus from the 'poor' aiming longer and earning more points per shot, and inferred attentional neglect from their greater borrowing. This paper re-analyzes that data and a large-scale replication by the same authors, then adds a new conceptual replication.
Why points-per-shot mixes two different play patterns
In the game, 'rich' participants had 15 shots per level and 'poor' participants had 3. Targets disappear when hit and are not replaced, so later shots on a level have fewer targets left to hit. Averaging points-per-shot across the whole game therefore compares 'poor' participants, who only ever took the first three shots of a level, with 'rich' participants, who also took many later, less profitable shots.
The re-analysis shows that later shots were significantly less profitable for all participants (β = −0.19, t = −14.85, p < .001). 'Poor' participants used significantly fewer shots per level than 'rich' participants (2.51 ± 0.45 versus 5.67 ± 1.79), so they simply did not take the low-yield shots that pulled the 'rich' average down. The higher points-per-shot of the 'poor' can arise from this structural difference without any difference in attention.
Borrowing followed opportunity, not neglect
The neglect claim rested on 'poor' participants borrowing a larger fraction of their budget than 'rich' participants (0.24 ± 0.15 versus 0.02 ± 0.05). But the groups did not face the borrowing decision equally often. 'Poor' participants encountered the decision on 45% of all levels, compared with 2% for the 'rich' (Χ2(1) = 184.42, p < 0.001), because their small budgets were exhausted sooner.
When the analysis is restricted to cases where borrowing was actually possible, the pattern reverses in direction: 'poor' participants were, if anything, more likely to resist borrowing (31%) than 'rich' participants (0%), though this was not significant (p = 0.052). Borrowing was also suboptimal for everyone: across both groups, performance was negatively related to borrowing frequency (r(33) = −0.688, p < 0.001). The authors argue that the apparent over-borrowing reflects how often the option appeared, not a failure to attend to its costs.
A new experiment tests attention to future information
The re-analysis is correlational in the sense that it reinterprets existing behavior, so the authors ran a conceptual replication with a direct test of neglect. They adapted the game so that, on the first 10 levels, participants could buy insurance against a 50% chance that a separate pineapple field would be destroyed. Eight of the insurances had positive expected value and two had negative expected value, so a participant attending to future outcomes should buy the good ones and avoid the bad ones.
If scarcity caused neglect of information outside the tunnel, 'poor' participants should have bought fewer positive-value insurances. They did not: purchase likelihood was 4.74 ± 2.84 for the 'poor' and 4.92 ± 2.92 for the 'rich' (F(1, 172) = 0.17, p = 0.682), and equivalence testing supported statistical equivalence given the pre-specified smallest effect size of interest. On negative-value 'scam' insurances, 'poor' participants were significantly less likely to buy than 'rich' participants (F(1, 172) = 8.24, p = 0.005), the opposite of what neglect would predict.
The replication reproduced the pattern and the artifact
The new experiment was not a failed manipulation. It reproduced the original behavioral signature: 'poor' participants took significantly more time aiming (F(1, 172) = 10.93, p = 0.001) and earned more points-per-shot across the game (F(1, 172) = 21.95, p < 0.001). That makes the null results on the shared shots and on insurance more informative, because the same study could produce the original effects.
It also reproduced the structural explanation. There was no meaningful difference in points-per-shot on the first three shots of each level (all ps > 0.439), and equivalence tests supported equivalence on the first, second, and third shots. Later shots were again significantly less profitable for all participants (β = −0.29, t = −34.28, p < 0.001). Because 'poor' participants were restricted to the first three shots, they again avoided the low-yield shots that lowered the 'rich' average.
What this means for the tunneling hypothesis
The paper's conclusion is narrower than 'scarcity has no effect on attention'. It contests the interpretation of the specific behavioral evidence taken to demonstrate tunneling, and it argues that the construct itself needs greater clarity. The authors note that other cited studies have shown greater focus under scarcity, such as better memory for prices or faster detection of scarcity-related words, but have not shown that this focus takes attention away from other beneficial information.
The practical priority is therefore to separate the two halves of the claim. A study that shows longer aiming, better recall, or faster detection supports increased focus; it does not by itself support neglect. Testing tunneling requires a measure of information outside the tunnel, such as the insurance choices here, and a design in which 'rich' and 'poor' participants face the same opportunities. The re-analysis shows that when the shared-shot comparison is used, the focus advantage disappears, and when insurance is offered, the predicted neglect does not appear.
Where the conclusion stops
The evidence is bounded by the game paradigm. Scarcity was manipulated as a budget of shots in a lab-style online game, not as lived poverty, and the samples were not representative of the U.S. population. The original study had N = 68 and the large-scale replication had a final sample of 974 after exclusions; the new experiment had 174 participants. The authors also note that the game program did not retain dropout information, so systematic dropout or incomplete randomization cannot be ruled out.
Within those bounds, the paper's specific judgment is that the 'Angry Blueberries' results should not be read as evidence of attentional tunneling. The more parsimonious account is that the game gave 'poor' and 'rich' participants different numbers of shots and different frequencies of the borrowing decision, and those structural differences produced the behavioral gap. Future tests of tunneling need a design that holds those opportunities equal and measures attention to information outside the scarce resource.
Sources used in this answer
Revisiting evidence for an attentional tunneling effect caused by scarcity.
Re-analysis of the 'Angry Blueberries' data and a new experiment show that 'rich'-'poor' behavioral differences are better explained by the game's structure than by a scarcity-induced attentional shift.