Pictures of nine high-calorie foods, including biscuits, chocolate and crisps, always fell into the "stop" category while images of healthy foods such as fruit and vegetables elicited a "go" response.
Other images of inedible items such as furniture and gardening tools could fall into either the "stop" or "go" bracket.
The idea was to train people to associate calorie-dense food with the act of "stopping".
Another group of 42 "control" volunteers played the same game, but only using non-food images that were not expected to influence their eating habits.
Lead scientist Dr Natalia Lawrence, from the University of Exeter, said: "These findings are among the first to suggest that a brief, simple computerised tool can change people's everyday eating behaviour.
"It is exciting to see the effects of our lab studies translate to the real world. This research is still in its infancy and the effects are modest - larger, registered trials with longer-term measures need to be conducted.
"However, our findings suggest that this cognitive training approach is worth pursuing. It is free, easy to do and 88 per cent of our participants said they would be happy to keep doing it and would recommend it to a friend. This opens up exciting possibilities for new behaviour change interventions based on underlying psychological processes."
Participants took part in the sessions at home or at work, and had to complete food diaries a week before and a week after training.
The results, published in the journal Appetite, showed that the energy intake of volunteers in the "active" group fell by 220 calories per day while no change was seen in the "control" group.
While "active" participants had lost 0.67kg on average a week after training, the average weight of those in the control group actually went up slightly by 0.17kg.
Participants varied in weight with Body Mass Index scores ranging from a healthy 21 to an obese 46.
- PAA
