Visual short-term memory load disrupts subitizing limit

Simone Cutini, Francesco Sella & Marco Zorzi

Department of General Psychology, University of Padova, Italy

simone.cutini@unipd.it

http://www.mendeley.com/profiles/simone-cutini/

ABSTRACT

Early visual processing is characterized by the parallel elaboration of a massive amount of information, nonetheless the number of objects that can be simultaneously

tracked and memorized is surprisingly limited, as observed both in visual short-term memory (VSTM) and non-symbolic visual enumeration tasks. Here we devised a

dual-task paradigm approach to address the nature and the selectivity of the link between VSTM and subitizing. Verbal memory load was used as control condition and

a stringent method to evaluate the individual subitizing range was employed. Our results demonstrate that VSTM load, but not verbal load, modulated performance

within the subitizing range without affecting the estimation of larger numerosities. This finding provides converging evidence regarding the presence of two distinct

mechanisms specifically associated to subitizing and estimation. Importantly, we found a striking correspondence between the number of elements retained in VSTM

and the decrement in the number of elements that can be subitized. In particular, the trade-off between VSTM load and enumeration accuracy at the subitizing limit

strongly suggests that VSTM and subitizing share the same cognitive resources.

METHOD

Participants: Twelve participants (5 males; Mean age = 23.6, SD = 3)

Verbal Memory Visual short-term memory

CONCLUSIONS

The present investigation aimed at unveiling the genuine nature of the relation between VSTM and subitizing. Our results demonstrate that the amount of VSTM load

modulates enumeration performance in the subitizing range without affecting the ability to estimate numerosity beyond the subitizing limit. Indeed, accuracy of enumeration

was affected by VSTM load only for numerosities around the subitizing limit (i.e., 3, 4 and 5 dots). This dissociation provides further evidence for the claim that subitizing

and estimation are related to two different mechanisms (e.g., Hyde, 2011; Piazza et al., 2011; Revkin et al., 2008) rather than being two extremes belonging to the same

continuum. Behavioral dissociation between subitizing and estimation was also highlighted in the study of Burr, Turi and Anobile (2010) as a function of attentional load

induced by dual task conditions (but see Vetter, Butterworth & Bahrami, 2008, for contrasting results). Moreover, different neural signatures for subitizing and estimation

have been highlighted in recent neuroimaging studies (Cutini et al., in press; Ester, Drew, Klee, Vogel, & Awh, 2012). More importantly, our results provide compelling

evidence with regard to the existence of a specific and selective link between VSTM and subitizing. We calculated the individual subitizing limit with a stringent procedure

and we found a remarkable correspondence between the number of elements retained in VSTM (as measured by Cowan’s K) and the decrement in the number of

elements that could be subitized. In particular, the trade-off between VSTM load and enumeration accuracy at the subitizing limit provides compelling evidence that VSTM

and subitizing share the same cognitive resources.

References: Burr et al. (2010). Journal of vision, 10, 1–10. Cutini et al. (in press). Neuroimage. Ester et al. (2012). The Journal of neuroscience, 32, 7169–77. Hyde, D. C. (2011). Frontiers in human neuroscience, 5, 150. Revkin

et al. (2008). Psychological science, 19, 607–14. Piazza et al. (2011). Cognition, 121, 147–53. Vetter et al. (2008). PloS one, 3, e3269.

RESULTS

http://ccnl.psy.unipd.it

We then retrieved for each participant the accuracy of enumeration at the individual

subitizing limit under the three different load conditions (VM, low-load VSTM, high-load

VSTM.

Enumeration accuracy at the individual subitizing limit, Repeated measure ANOVA with:

Load Type (Verbal, VSTM-Low, VSTM-High): F(2, 11) = 8.16, p = .002

Enumeration Accuracy (black line):

VM>VSTM-Low load (p = .035); VSTM-Low load>VSTM-High load (p = .022).

Cowan’s K (i.e., the number of items memorized in the VSTM task, gray line) was

significantly lower in the low-load VSTM (M = 0.82, SD = 0.11) as compared to the high-

load VSTM condition (M = 1.65, SD = .7), t(11) = 4.26, p = .01

Participants accomplished two different memory tasks

in a dual-task condition.

Verbal Memory (VM) task: four disyllabic pseudo-words

were presented through earphones for four seconds (1

second each).

low-load condition: one word repeated four times;

high-load condition: four different words.

VSTM task: a 2 x 2 memory array with four figures was

presented for 500 ms.

low-load condition: the 4 figures were identical;

high-load condition: the 4 figures were all different.

After the memory stimuli, a cloud of dots (numerosity

from 1 to 9) was shown for 200 ms and then replaced by

a mask that lasted for 100 ms.

Task 1: report the number of dots (keyboard).

Task 2:judge whether the test figure/word was present in

the memory array (Y/N).

Dots enumeration

VSTM load–subitizing trade-off

Omnibus ANOVA on dots enumeration accuracy: All main effects

and two-way interactions were significant; Memory Task x Load x

Numerosity, F(8,11) = 2.10, p = .044, indicating that dots numerosity and

load amount had a different impact depending on load type.

Accuracy in the low-load vs. high-load conditions for each numerosity

enumeration performance was worse in the high-load condition only for

numerosities around the subitizing limit (3, 4 and 5 dots, p<.05) .

Results suggest an Interplay between VSTM and subitizing.