Modulating Attentional Load Affects Numerosity Estimation: Evidence against a Pre-Attentive Subitizing Mechanism

Bahrami B (2008) Modulating Attentional Load Affects Numerosity Estimation: Evidence against a Pre-Attentive Subitizing Mechanism. PLoS ONE 3(9): e3269. doi:10.1371/journal.pone.0003269 Modulating Attentional Load Affects Numerosity Estimation: Evidence against a Pre-Attentive Subitizing Mechanism Petra Vetter 0 Brian Butterworth 0 Bahador Bahrami 0 Eric Warrant, Lund University, Sweden 0 1 Institute of Cognitive Neuroscience, University College London , London , United Kingdom , 2 Department of Psychology, University College London , London , United Kingdom Traditionally, the visual enumeration of a small number of items (1 to about 4), referred to as subitizing, has been thought of as a parallel and pre-attentive process and functionally different from the serial attentive enumeration of larger numerosities. We tested this hypothesis by employing a dual task paradigm that systematically manipulated the attentional resources available to an enumeration task. Enumeration accuracy for small numerosities was severely decreased as more attentional resources were taken away from the numerical task, challenging the traditionally held notion of subitizing as a pre-attentive, capacity-independent process. Judgement of larger numerosities was also affected by dual task conditions and attentional load. These results challenge the proposal that small numerosities are enumerated by a mechanism separate from large numerosities and support the idea of a single, attention-demanding enumeration mechanism. - Funding: This work was supported by a European Marie Curie Research Training Network grant to Butterworth, a Marie Curie Early Stage Research Fellowship to Vetter and a UCL Graduate School Research Scholarship and an Overseas Research Scholarship to Bahrami. Competing Interests: The authors have declared that no competing interests exist. Jevons found that he could estimate the number of beans in a box without error when there were four or fewer, but became increasingly inaccurate as the number of beans increased beyond four [1]. Subsequent studies have confirmed his findings, and it is now generally assumed that the immediate and accurate apprehension of the numerosity of collections of four or fewer objects uses a process separate from enumerating larger collections [26]. Following Kaufman and colleagues, this process is called subitizing [7]. The current basis for this distinction has come from a discontinuity in the slope of the curve that relates enumeration time to the number of items to be enumerated. Enumeration in the subitizing range (1 to 3 or 4 items) typically yields a shallow slope whereas the slope for 5 items and above (the counting range) is considerable steeper. This pattern has traditionally been fitted with a bilinear function and two functionally separate enumeration mechanisms have been inferred (see [2] for a review). Furthermore, by analogy with classical studies of visual search [8], a parallel and pre-attentive process has been inferred from the shallow slope for subitizing (equivalent to pop-out search) and a serial and attentive process from the steeper slope (equivalent to conjunction search) for counting [3,9]. Support for this distinction has come from brain imaging studies that show quantitative differences in parietal lobe activity for the counting range as compared with the subitizing range [4,6]. More specific evidence for a pre-attentive subitizing mechanism has come from a neuropsychological study of neglect patients [10]. Neglect patients with extinction, who cannot report items in the contra-lesional field due to their inability to attend to this side of space, can nevertheless enumerate up to four objects when two of them are in the neglected field [10]. However, one brain-imaging study has failed to distinguish between the neural substrates of subitizing and counting, and found instead that human parietal cortex activation increased linearly with the number of items [11]. Balakrishnan and Ashby questioned the basis of the initial inference of two mechanisms from the performance data by demonstrating that a bilinear fit is unjustified and a continuous model of enumeration is equally supported by the performance data [12,13]. Moreover, the strong notion of pre-attentive/attentive dichotomy has been regarded as an oversimplified account in the attention literature (e.g. [14,15]) and particularly the hypothesis of attention-free perceptual processing has been questioned [16,17]. Indeed there is evidence that even the simplest forms of feature detection (e.g. orientation detection), which had previously been thought of as occurring pre-attentively, depend on the availability of attentional resources in a dual-task situation [18]. In this study, we investigated how the judgement of both small and large numerosities is affected by a withdrawal of attentional resources, and more specifically, we tested the hypothesis that subitizing is a pre-attentive process. We reasoned that if subitizing is pre-attentive, it should be unaffected by experimental manipulations such as dual-task paradigms that reduce the availability of attentional resources [16,18]. In addition to imposing an additional task onto a numerosity judgement task, we employed the framework of load theory [19,20]. Load theory states that in a dual task situation, processing of secondary task stimuli depends on the attentional requirements of the primary task. Under high attentional load, processing capacity is entirely dedicated to the primary task leading to reduced (and sometimes eliminated) processing of the secondary task. Under low attentional load, however, the capacity limit is not reached and attentional resources spill over to perform the secondary task. In this experiment, we combined a secondary numerosity judgement task with a primary task with two levels of attentional load (low and high load). We predicted that if subitizing is a preattentive process, it should not be affected by dual versus single task manipulations and, more importantly, subitizing should not be affected by attentional load. However, if subitizing is constrained by attentional capacity, it should be compromised by both experimental manipulations. Subjects 14 subjects (mean age: 23.1, 10 females) with normal or correctedto-normal vision participated. All gave written informed consent and were paid for their participation. The study was approved by the ethics committee of the Dept. of Psychology at UCL. Visual stimulus The visual stimulus consisted of: (i) a central diamond shape (4u of visual angle) comprising 4 coloured triangles and (ii) a circle of gabor patches (10u) on a grey background (see example stimulus in Fig. 1a). Eight different colour combinations were used for the central diamond shape (Fig. 1b). The gabor patches (2u each) in the circle were either vertically oriented high-contrast (100%) targets or horizontally orientated low-contrast (50%) distractors. The distance b (...truncated)