Labelling errors on fresh produce are estimated to cost the UK supermarket industry £50m per year in product recalls and wastage. Such errors occur despite robust quality control procedures. Given the financial and environmental impact of these errors, it is important to understand whether labelchecking performance can be predicted by individual differences in cognitive abilities. To this end, participants carried out a simulated label-checking task together with a number of measures of information processing speed, attention, short-term/working memory, and mind-wandering. Accuracy of label checking was found to be significantly predicted by three of the measures, with better short-term verbal memory being most strongly associated with performance. Cognitive tests such as these provide a means of identifying how well employees are likely to perform when undertaking such tasks and, if necessary, how they should be supported in that role, possibly forming a screening battery when recruiting new quality control staff. The findings highlight the importance of determining the component processes of cognition which contribute to performance in real-world work environments.
A long-standing concern of applied psychology has been to
provide the practical means by which to predict how well
individuals are likely to perform in real-world situations
along with a theoretical understanding of why this should be
the case. Indeed, the motivation for developing the first tests
of intelligence was not just to measure individual
differences but to assist in the appropriate placement of
individuals on the basis of their ability and likely
achievement
With advances in the study of cognitive psychology, it has
become clear that behavior relies on a variety of specific
and qualitatively different resources, each dedicated to a
different kind or aspect of processing
The measurement of relevant specific cognitive abilities,
such as the speed of information processing, the ability to
direct and sustain attention, the capacity to hold and update
information in memory, and the executive functions
necessary to plan and execute behavior
The primary challenge for research in this area is to provide a reliable basis for matching the particular cognitive skills of individuals with the demands of tasks they are, or will be, called on to perform. Clearly there are broad benefits in terms of recruitment, retention, morale and quality of performance in ensuring that employees are given work that suits their particular competencies. Failing to do so will almost certainly lead to poorer performance, and depending on the role in question, may have high financial implications or costs in terms of ill-health, injury or even death.
Advances in understanding the role of specific cognitive
abilities in task performance also promise to reduce ethnic
and cultural biases that occur when general mental ability is
used as the sole basis for employee selection, assignment.
Such biases are likely to reduce the chances of individuals
with disabilities gaining employment, even though they
might be shown to be perfectly able to undertake the job if
relevant specific cognitive abilities had been assessed. This
may be the case, for example, for some individuals with
autism who have a normal or even superior ability to attend
to detail, even though they may be deficient in other aspects
of cognition
There are, therefore, compelling theoretical and practical reasons to pursue research that promises to provide both a better understanding of the cognitive abilities that particular kinds of tasks require and to map these onto specific abilities individuals possess. Such matching would optimize the performance of both the individual and the system in which he or she works.
The research reported in this paper investigated whether scores on different tests of specific cognitive processes could predict the accuracy of performance on a repetitive label checking task. This task was designed to closely resemble work that is undertaken by quality control inspectors at a fresh produce packaging facility in the UK. Measures of visual search, perceptual speed, short-term memory, and attention were administered, together with a self-report measure probing the propensity of individuals to mind-wandering during ongoing behaviour.
The label-checking procedure involves an operative determining whether or not the information that appears on a given product label correctly matches details as set out on the product specification sheet (which includes information about the supermarket’s order as well as the product from the producer). The number of fields of information printed on a label varies between three and eleven. Example fields are the name of the product, its weight, its country of origin and its barcode. If the information which appears on the product label does not match the specification sheet, the quality control checker should detect this and reject the label. Generally three or four independent quality control checks are performed before the order is shipped from the packaging facility to supermarket distribution depots.
Despite these stringent quality control procedures, products that are erroneously labelled do sometimes escape the packaging facility, necessitating the recall and disposal or repackaging of produce. The recall and disposal of food due to label errors is estimated to be £50 million industrywide annually in the UK alone (S. Hinks, Product Technical Manager: Fruit and Floral, Sainsbury’s Supermarkets Ltd, personal communication). Whilst infrequent, the financial and environmental costs attached to label errors are such as to drive research into their reduction.
Given the accuracy-driven and time-constrained work
environment in which label-checking occurs, two different
measures of the speed and accuracy with which information
could be processed were administered. Visual search tasks
Short-term memory relates to the ability to store
information temporarily in memory over a duration of
seconds
The Attention Network Test
Mind-wandering occurs when an individual has thoughts
unrelated to the task which move attention from the
intended task. The Daydreaming Frequency Subscale
Together, the battery of tests was designed to measure a broad range of specific cognitive functions that might underpin and predict performance on label checking and other quality control tasks that require the identification of mismatches or mistakes.
A total of 51 university students (44 females, 7 males, mean age = 24 years, SD = 6) took part in the experiment. They received a small honorarium or course credit in appreciation of their participation. All of the participants reported themselves to be naïve to the quality control processes involved in checking fresh produce labels.
The participants were either native English speakers or were studying at undergraduate degree level with an International English Language Testing System (IELTS) score of at least 6.0 (the minimum requirement of London South Bank University for entry to its degree courses).
The label-checking and visual search tasks were programmed and run in Experimenter Builder Version 1.4.128 B (SR Research Ltd., Ontario, Canada). E-Prime 2.0 (Psychology Software Tools, Inc., Sharpsburg, PA) was used to program and implement the remaining computerized tasks.
Facsimiles of the product specification sheet and labels used in the packaging facility were created for the purpose of the experiment (Figures 1 and 2 respectively). The number of fields of information per product on the specification sheets and produce labels was held constant at seven. These fields of information were the product (the type of fruit or vegetable, e.g., baby courgettes), country of origin, the grower (the name of the company which grew and shipped the product), the quantity of items contained in the packet (i.e., the weight of the product), its best-before date (indicated by “BB” on the specification sheet), the product’s barcode number, and details of any promotion ribbon or label to be appended to the packaging (i.e., any promotional activity on the product being offered by the supermarket, such as “Any 2 for £2.50”). In the course of the block of trials, fifty different labels were presented.
The produce label and the product specification sheet were presented simultaneously on a 21”colour monitor screen, with the former occupying the top half and the latter the lower half of the display.
A head-rest was used in the label-checking task in order to minimize the head movements of the participants.
A block of 50 trials was presented. The information displayed on the product specification sheet and that presented on the label matched on 40 of these trials. For the remaining 10, there was a mismatch between the two sources of information. For each trial where there was a mismatch, only one field of information varied between the product specification sheet and the produce label (e.g., the best-before date). The field of information that differed was varied pseudo-randomly over the 10 trials such that the errors appeared in different fields. Responses to these trials were logged as correct when a mismatch between the information set out on the product specification sheet and the label was indicated by the participant.
The participants undertook two further 50-trial labelchecking blocks after this initial block. The data relating to these are reported in Smith-Spark, Katz, Marchant, and Wilcockson (2015). The focus of the current paper, however, was purely on the extent to which the initial labelchecking performance of individuals with no prior experience or training could be predicted on the basis of scores from the battery of cognitive tasks which was administered to them.
Visual search ability was measured using a modified version of Triesman and Souther’s (1985) letter finding task. Participants were presented with an array of 19 letter stimuli (namely, N, C, F, K, and P). In one block of trials, they were asked to locate a normal, forward-facing letter in an array of backwards, mirrored letters. In a separate block of trials, the participants were asked to identify a backwards letter amongst an array of normal, forward-facing letters. In each case 1, 2, or 3 letters faced in the opposite direction to the others. Participants were asked to indicate how many backwards-facing letters they had seen. Performance on the backwards and forwards trials was combined to give mean RT and accuracy scores for visual search ability.
Perceptual speed was measured using a letter comparison
task, modified from
Phonological short-term memory was assessed by the Digit Span Task. Participants were presented with a sequence of single digit numbers, one at a time. Once the sequence was completed, they were asked to recall the digits in the order they had been presented. The number of digits gradually increased over trials, starting with two and going up to a maximum of 10. Three trials were presented at each level. At least two of the three trials needed to be correct in order to advance to the next level of the task. A participant’s span length was taken as the last level at which they could reliably remember the sequence of digits in the correct serial order. A backward digit span task was also administered in which participants had to report the digits in reverse serial order, thereby drawing on working memory rather than simply short-term memory to store and manipulate information simultaneously.
The Corsi Block span test
A modified version of the Visual Patterns Test
The ANT
The Daydreaming Frequency subscale (DFS) of the
Imaginal Process Inventory
A multiple stepwise regression was run with the cognitive test measures entered as predictor variables. Overall labelchecking accuracy was the outcome variable.
Informed consent was given by all participants to take part in the experiment. Before the checking task began, the participants were seated at a viewing distance of 55cm from a 21” computer monitor. They then viewed a 10-minute slide show presentation. This provided them with a detailed description of the label layout, specification sheet layout, general task instructions, the nature of errors, etcetera.
During the label-checking task, the participants indicated whether or not the information presented on a given label was correct, checking it against the appropriate entry on the specification sheet. They were instructed to respond as quickly but as accurately as possible. Responses were made by pressing designated Yes and No keys on a standard QWERTY keyboard.
The cognitive measures were administered in a separate testing session. The order in which the cognitive tasks were presented was counterbalanced between participants. The letter and number comparison tasks had a pen-and-paper format, while all others were computerized.
The participants were debriefed upon completing testing.
The scores from three participants were removed on the backward search and two on the forwards search due to their having mean scores more than 2.5 SDs from the overall mean.
The overall mean proportion accuracy of label-checking was .85 (SD = 0.05).
Descriptive statistics for each cognitive test are displayed in Table 1, together with Pearson’s correlations indicating the extent of the relationship between each test and label checking accuracy.
The stepwise multiple regression analysis indicated that overall label-checking accuracy could be significantly predicted by the cognitive predictors, R = .637, adjusted- R2 = .358, F(3, 37) = 8.44, p < .001. Three predictors were entered in the final three-step model. These were digit span forwards, standardized-β = .658, p < .001, Corsi forwards, standardized-β = -.395, p = .004, and perceptual speed, standardized-β = -.459, p = .004.
The simulated label checking task used in this study resulted in a rate of errors somewhat greater than that indicated by the historical record at the actual packing facility on which it was modelled (approximately 15% as opposed to 2% of checks). While the stimuli were virtually identical, the laboratory-based task did entail many more checks and in a more concentrated time-frame than demanded in this and most likely other real-world situations.
The results indicate that label-checking accuracy can be
significantly predicted on the basis of the cognitive tasks
employed in this experiment. Verbal short-term memory (as
measured by the digit span forwards task) was the strongest
predictor of performance, with the ability to retain a larger
number of digits in memory being associated with higher
accuracy. The next strongest predictor was perceptual speed
although, in this case, the relationship was negative. It
would appear that processing information more rapidly was
associated with lower accuracy, which may indicate a
speed-accuracy trade-off. Spatial short-term memory
(measured by the Corsi forwards task) was also a significant
negative predictor of accuracy. Although it may seem
paradoxical that the ability to hold more spatial information
in memory would be associated with poorer accuracy, it
may be that a stronger spatial memory encouraged
individuals to adopt a non-optimal approach to
labelchecking, in particular chunking
Whilst null results should be treated with caution, the results suggest that cognitive tasks involving greater executive resources do not predict performance, since none of the executive-loaded span tasks were significantly associated with label-checking accuracy. Further to this, neither visual search abilities nor the ANT predicted performance, suggesting that neither visual search nor the attentional processes tested by the ANT contribute to labelchecking accuracy. Finally, mind-wandering, as measured by the DFS), did not predict correct responses on the labelchecking task.
The present study explored the value of tests of specific
cognitive functions as predictors of performance on a
simulated label checking task. Unlike most research in
applied areas of occupational psychology, this experiment
had well defined, objective outcome measures and allowed a
reasonably close mapping between the behavioural
requirements of the task, i.e., perceptual scanning,
comparison, no problem solving, etc., with narrowly defined
cognitive processes which one would assume underpinned
these actions, such as visual search, focused attention,
executive control and short-term memory. While some
success in prediction was gained, the experiment also
demonstrated the challenge in determining the connection
between specific cognitive abilities and task performance.
This is partly because there are different ways in which a
given task, even a relatively straight-forward task like label
checking, can be approached
Aside from the strength and availability of specific
cognitive resources, some of which have been measured in
the current experiment, performance also depends on the
demands of situational factors such as time constraint,
interruptions, incentives and cognitive load, and as
importantly non-cognitive factors such as previous
experience, motivation and conscientiousness. Together
these lead to cognitive dynamics which are variable and
difficult to predict, as seen in the negative contribution of
spatial memory and processing speed to the accuracy of
performance. Given the manifold nature of cognition, even
basic procedural tasks such as label checking, may resist an
exhaustive description of the contribution of specific
cognitive processes to performance. This is probably why
tests of general cognitive ability have generally proven to be
superior predictors of job performance as well as the
preferred basis for employee selection and allocation
The research reported in this paper was funded by Innovate UK (grant number 101393). The authors are very grateful to Simon Hinks (Sainsbury’s Supermarkets Limited), Daniel Boakes (Mack), Tetyana Bennett (Mack), Trish Fox (Mack), and Jez Pile (Muddy Boots Software). The authors also thank Monika Michalska for assistance with data collection and our grant monitoring officer, John Stones, for support.