Temporal characteristics of working memory for spatial relations: An ERP study

https://doi.org/10.1016/j.ijpsycho.2010.04.008Get rights and content

Abstract

Spatial relations can be represented categorically, by means of abstract labels, or coordinately, in metric, absolute measures. These representations have been associated to the left and the right hemispheres respectively (Kosslyn, 1987). Recent studies have focused on the temporal dynamics of spatial relation processing, with working memory task designs. In this light, we examined the suggested lateralization effect in an ERP study incorporating a visual half field match-to-sample design, in which two sequentially presented stimuli were compared. By manipulating the length of the retention intervals between the two stimuli (500 ms, 2000 ms, and 5000 ms), spatial working memory effects were studied at three separate stages of working memory; encoding, memorization, and retrieval. The hypothesized interaction of instruction and visual field was found in the behavioural data, restricted to the 2000 ms retention interval. The EEG data indicate a strong overall right hemisphere effect, which is likely related to spatial working memory in general. Categorical and coordinate processing appears to already differentiate during the encoding stage in the P300 complex (300–500 ms after presentation of the first stimulus), where instruction interacts significantly with hemisphere in the parietal area. We found a clear right hemisphere advantage for coordinate processing and no lateralization for categorical processing. We argue that the outcome indicates qualitative rather than quantitative differences between categorical and coordinate processing.

Introduction

The spatial relations we encounter in our environment can be divided into two types: categorical and coordinate relations. Categorical spatial relations concern abstract, qualitative relations such as “above”, or “left of”, whereas coordinate spatial relations are metric and quantitative in nature, describing exact distances between objects (Kosslyn, 1987). The mechanisms underlying the processing of these two types of relations are commonly thought to be dissociable at the neural level. A left hemisphere advantage is often found for categorical processing, while the right hemisphere is clearly linked to coordinate processing (e.g. Kosslyn et al., 1989). This lateralization effect has been replicated in numerous studies, most of which concerned behavioural half field experiments (for a review see Jager and Postma, 2003). Others have provided evidence for lateralized processing in neuropsychological studies (Laeng, 1994, Van Asselen et al., 2008, Palermo et al., 2008). Remarkably, the number of neurophysiological and imaging studies directly addressing the distinction is limited. Thus far, three functional imaging studies have substantiated the foregoing (Baciu et al., 1999, Trojano et al., 2002, Van der Ham et al., 2007), but critical findings have also been reported (Martin et al., 2008).

Importantly, an increasing number of studies have applied a working memory design to study spatial relation processing (e.g. Laeng and Peters, 1995, Van der Ham et al., 2007, Martin et al., 2008, Van der Ham et al., 2009), instead of a single stimulus perceptual design (e.g. Baciu et al., 1999). Furthermore, in a recent fMRI study (Van der Ham et al., 2009) we observed a clear double dissociation in the superior parietal cortex, only with a 2000 ms interval between two stimuli in a single, match-to-sample trial, and not with a 500 ms interval. These results indicate that temporal characteristics of spatial relations processing should be taken into account when studying lateralization patterns as the duration between the first and second stimulus appears to be an influential temporal factor. A relevant question in this light is whether categorical and coordinate relation processing differences emerge very early after stimulus presentation, or at a later processing stage. A proper means to examine these temporal patterns is EEG measurements, with its high temporal resolution. It allows us to infer how categorical and coordinate spatial relation processing differ for each of the underlying cognitive mechanisms that occur within a single trial. Most of the studies so far have only provided behavioural outcomes or imaging data that represent a combined outcome of all cognitive processes that occur within a trial. The behavioural data are likely to primarily reflect the retrieval and comparison process, because only the second stimulus is presented laterally and requires a response, while the imaging data provide the activation patterns averaged for a complete trial and therefore include all processes occurring during a single trial.

As far as we know, only one full EEG study has been published focusing on the differences between categorical and coordinate relation processing (Van der Lubbe et al., 2006). In this study such a match-to-sample design was used, similar to the design in our previous studies (Van der Ham et al., 2007, Van der Ham et al., 2009). One stimulus (S1) presented centrally was compared either categorically or coordinately to a second stimulus (S2) presented laterally. This format combined with EEG measurements enabled Van der Lubbe et al. (2006) to examine the differences between categorical and coordinate spatial processing at different processing stages, which represent the different cognitive mechanisms underlying the task. In a single trial three separate stages were identified: a) the encoding stage related to the encoding of the first stimulus, b) the memorization stage in which the stimulus is retained until the second stimulus appears, and c) the retrieval stage, in which the relevant properties of the first stimulus are retrieved to compare them to the second stimulus. Van der Lubbe et al. (2006) mainly found quantitative and not qualitative differences between categorical and coordinate tasks and concluded that Kosslyn's original lateralization account could not be confirmed. During the encoding stage early differences (160–320 ms) between the tasks were found, but dipole analyses did not show the hypothesized categorical left hemisphere and coordinate right hemisphere advantage. In the memorization stage, there was more posterior right hemisphere activity for the coordinate task, compared to the categorical task, which is likely related to spatial working memory activity. In the retrieval stage early effects (120–240 ms) confirmed the visual half field technique and task dependent effects emerged around 400 ms after S2 presentation. As these effects are relatively late, they are suggested to reflect retrieval of S1 or judgment on the comparison, not the encoding of S2, and indicate stronger right hemisphere activity for the categorical task, compared to the coordinate task. Based on these specific time intervals and more general findings about stimulus discrimination and working memory tasks, the N2 and P300 complexes seem of particular relevance in our study. N2 reflects the detection of stimuli and its amplitude increases as stimulus discrimination becomes more difficult (see e.g. Naatanen, 1986). P300 responses have been shown to reflect a number of processes all entailed in working memory (see e.g. Ellwanger et al., 1999, Shucard et al., 2009). Therefore we will examine the N2 and P300 complexes for both the encoding and retrieval stages, linked to the presentation of S1 and S2 respectively. To complement these analyses we will also make use of current source density estimations of these complexes, to gain more information about cortical activation during these critical time periods.

The memorization stage between the two stimuli is of specific interest, since when it is long enough (i.e. the 2000 ms and 5000 ms intervals), slow wave activity can be observed, which is commonly related to working memory processes (e.g. McEvoy et al., 1998, Awh et al., 1998, Ruchkin et al., 1997). For instance, a previous report showed an increase in negative slow wave activity at electrodes at the posterior parietal and occipital locations during working memory maintenance (Mecklinger and Pfeifer, 1996). In addition van der Lubbe et al. (2006) found task related effects 1000–2000 ms after S1 presentation, which they interpret as a reflection of spatial working memory activity.

The experimental design that we employed, provided two main advantages with regard to the designs previously used to test categorical and coordinate relations (see Jager and Postma, 2003). We have varied the length of the retention interval as we have done in our previous studies (Van der Ham et al., 2007, Van der Ham et al., 2009). Previous empirical findings and theoretical inferences (e.g. Kosslyn, 1987, Huttenlocher et al., 1991) suggest that categorical and coordinate spatial relation processing differs with longer intervals, i.e. with a larger load on working memory. As categorical spatial relations can serve a communicative purpose, and are relatively easily memorized by means of prepositional terms, they show little decay with longer retention intervals. In contrast, coordinate spatial relations are mainly used for instantaneous actions like grasping and navigation, which require no elaborate memorization. The selected intervals of 500, 2000, and 5000 ms have been shown to adequately represent these differential patterns of decay in memory (Postma et al., 2006, Van der Ham et al., 2007). Furthermore, by varying the duration of the retention intervals we can address the issue of strategy use at a neural level. For instance, in order to retain a visual stimulus in memory longer, one could recode the visual image verbally, offering a more persistent memory code. Such transformations have been reported by subjects when asked about their strategy use (Van der Ham et al., 2007, Van der Ham and Postma, in press). The tasks were solved in a more perceptual manner in the shorter retention intervals while verbal strategies were more frequently used with longer intervals. As these strategy effects are only based on self-reports, more objective data related to strategy use and working memory processes are required to confirm these suggestions. If verbalization is the preferred strategy for some specific conditions, we would expect activity in language related areas, such as Broca's area.

The second advantage of the current paradigm is a change in stimulus characteristics, which in our opinion is an improvement over the stimuli used by van der Lubbe et al. (2006). We applied the cross–dot stimuli, instead of the often used dot–bar stimuli. The cross–dot stimuli overcome the problem of the dichotomous comparisons of above/below and near/far, which has commonly resulted in practice effects for coordinate trials (e.g. Baciu et al., 1999). We have doubled these possibilities by constructing four different categories and four different distances. As subjects have shown to be unable to deduct the pattern of possible dot positions, we can avoid problems like practice effects, in which subjects might assign categories to the coordinate positions. In addition, it somewhat increases the level of difficulty of categorical trials, reducing the difference in difficulty between categorical and coordinate trials.

In short, the aim of the current study was to further examine the processing and lateralization of categorical and coordinate spatial representations and their temporal properties. We expected to find the hypothesized lateralization pattern for the shorter intervals, congruent with previous findings. The potential use of a more verbal strategy, as opposed to a visuospatial strategy, was expected to emerge for the longer intervals and categorical processing in particular.

Section snippets

Participants

Twenty-four subjects participated in the study in exchange for course credit or money. Five participants were discarded from the dataset; four of them showed behavioural results at chance level in at least one of the conditions, the fifth one had low quality EEG recordings. The data presented here originate from the remaining nineteen participants, 10 male (mean age: 22.00, SD: 2.79) and 9 female (mean age: 21.89, SD: 2.52). All participants were unaware of the rationale of the experiment, had

Behavioural results

Mean RTs and ERs in all conditions are shown in Table 1 and Fig. 4. For ERs there was a significant main effect for instruction, F(1,18) = 65.05, p < 0.001, and visual field, F(1,18) = 7.01, p = 0.016. ERs were higher in case of a coordinate instruction and when S2 was presented in the RVF/LH. There were significant interaction effects for instruction  retention interval, F(2,17) = 13.02, p < 0.001, retention interval  visual field, F(2,17) = 5.21, p = 0.017., and instruction  retention interval  visual field, F

Discussion

The aim of the present study was to use EEG measurements to examine the temporal characteristics of the processing of categorical and coordinate spatial relations and the role of spatial working memory mechanisms. A match-to-sample design combined with the temporal resolution of EEG measurements enabled us to examine three separate stages of information processing comprehensively. The ERP signal related to S1 presentation reflects the encoding stage, the memorization stage starts after S1

Conclusion

The expected coordinate right hemisphere advantage and the categorical left hemisphere advantage were confirmed by the behavioural as well as the EEG data. The behavioural effect was limited to the intermediate retention interval and the categorical lateralization was weaker, concurring with previous empirical findings. In the EEG data, the interaction of instruction and hemisphere was also found with a strong right hemisphere advantage for coordinate processing that was absent for categorical

Acknowledgement

This study was made possible by a grant of the Netherlands Organisation for Scientific Research (NWO) (Evolution and Behaviour: 051-14-027).

References (34)

Cited by (13)

  • Effects of mnemonic load on cortical activity during visual working memory: Linking ongoing brain activity with evoked responses

    2013, International Journal of Psychophysiology
    Citation Excerpt :

    The late slow wave was predominantly confined to parietal channels. Based on the maximal deflections of the ERPs during encoding and retrieval, three channels over occipital (PO8), parietal (Pz) and frontal (FCz) areas were selected for further analysis (see scalp topologies in Fig. 4, and also van der Ham et al., 2010). Four peaks were selected and event-related activity was averaged over an interval surrounding those peaks (P1: 100–160 ms, N2: 180–260 ms, P3: 300–400 ms, slow wave: 400–1000 ms).

  • The effect of stimulus features on working memory of categorical and coordinate spatial relations in patients with unilateral brain damage

    2012, Cortex
    Citation Excerpt :

    This indicates that although there is no clear overall effect, there is a double dissociation of categorical and coordinate processing and LH and RH performance, respectively, but only when the stimuli with a small cross were presented. This is a confirmation of Kosslyn’s original proposal, and also in line with previous findings for the cross dot task with a small cross (Van der Ham et al., 2009; Van der Ham et al., 2010). Also the relative weaker impairment of the LH group for the categorical task is in line with previous findings stating that the LH involvement in categorical processing is often weaker or not present, whereas the coordinate RH involvement is clearly present (see e.g., Van der Ham and Postma, 2010).

  • Hemispheric differences in spatial relation processing in a scene perception task: A neuropsychological study

    2011, Neuropsychologia
    Citation Excerpt :

    Categorical processing was thought to show a left hemisphere advantage, whereas the right hemisphere would predominate in processing coordinate information. In many behavioural (e.g. Hellige & Michimata, 1989; Laeng & Peters, 1995; van der Ham, van Wezel, Oleksiak, & Postma, 2007) and neurofunctional studies (e.g. Baciu et al., 1999; Trojano, Conson, Maffei, & Grossi, 2006; van der Ham, Raemaekers, van Wezel, Oleksiak, & Postma, 2009; van der Ham, van Strien, Oleksiak, van Wezel, & Postma, 2010), this lateralisation pattern has been found for tasks testing categorical and coordinate relation processing. Neuropsychological studies thus far have been sparse but have also found supportive evidence (e.g. Laeng, 1994; Palermo, Bureca, Matano, & Guariglia, 2008; van Asselen, Kessels, Kappelle, & Postma, 2008).

  • A Study for Evaluations of Automobile Digital Dashboard Layouts Based on Cognition Electroencephalogram

    2021, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
View all citing articles on Scopus
View full text