183
A. Vincent et al. / International Journal of Psychophysiology 23 (1996) 181-198
Back to Psychophysiological / Home PageInternational Journal of Psychophysiology 23 (1996) 18I-198
Relations_among96.doc
Relations among memory performance, mental workload and cardiovascular responses
Alex Vincent a*, Fergus I.M. Craik b, John J. Furedy b
a Ergonomics Division, Transport Canada, Canada Building, 344 Slater Street, Ottawa, Ont. KIA ONS, Canada Department of Psychology, University of Toronto, Toronto, Canada
Received 1 February 1996; revised 25 June 1996; accepted 23 July 1996
Abstract
The levels of processing paradigm has been a powerful research framework in the study of memory for close to a quarter century. However, an objective index of depth of processing is still lacking. Two experiments using lists of words, presented to male subjects, were performed to compare the effects of depth of processing, rate of presentation, and task incentive on recognition memory performance, self-reported workload, and cardiovascular responding. Memory performance results from the two experiments demonstrated higher recognition levels associated with deeper processing and slower presentation rates. Deeply encoded items were associated with faster recognition latencies. Self-reported workload levels were higher for deeper processing and faster presentation rates. Cardiovascular responses were generally amplified with the addition of a task incentive. Increased blood pressure was associated with faster presentation rates. Increased heart rate and decreased T-wave amplitude (i.e., increased sympathetic activity) were uniquely associated with the deep encoding of information presented at the fastest rate. This particular encoding condition was associated with increased recognition levels. Deeply encoded items were associated with increased suppression of heart rate variability during recognition. This combination of behavioral and cardiovascular measures may provide the basis for an objective index of depth of processing.
183
A. Vincent et al. / International Journal of Psychophysiology 23 (1996) 181-198
1. Introduction
It is well established that different types of processing at the time of acquisition are associated with large differences in later memory performance. These relations have been described and studied under the heading of levels of processing. The general idea is that material encoded semantically is better recalled than material encoded in terms of its 'shallower' phonological or visual features (Craik and Lockhart,
* Corresponding author. e-mail: ; Fax: +1 (613) 9902913.
1972; Craik and Tulving, 1975). The levels of processing framework gives a good account of a number of important findings in memory research, but it also suffers from several limitations, among them the absence of an independent index of depth of processing, and whether the effects of processing depth simply represent a by-product of differences in effort (Ellis et al., 1984; Tyler et al., 1979). Craik and Lockhart (1972) proposed that memory performance reflected the depth, that is the degree of semantic analysis, to which encoded events were processed. However, Baddeley (1978) and others have pointed out the inherent circularity in the levels of processing formulation. In the absence of an independent index
183
A. Vincent et al. / International Journal of Psychophysiology 23 (1996) 181-198
0167-8760/96/$15.00 Copyright 0 1996 Elsevier Science B.V.
All rights reserved. PII SO167-8760(96)00058-X
183
A. Vincent et al. / International Journal of Psychophysiology 23 (1996) 181-198
of 'depth', it is too easy to claim that any well-remembered event must have been processed deeply, simply because of the high level of memory performance.
Psychophysiological and neuroimaging measures have the potential to break this circularity. If one or more of such measures can be shown to relate both to the degree of semantic involvement during the encoding process, and to the level of memory performance, it would put the concept of depth of processing on a more objective and rigorous footing. In fact, at least two studies have reported that differences in depth of processing correlate with event-related potentials (Ferlazzo et al., 1993; Sanquist et al., 1980). Sanquist et al. studied the sensitivity of event-related potentials (ERPs) to depth of processing by measuring the ERPs to different encoding tasks varying in processing level and to a subsequent test of recognition memory. The ERPs differed among the three encoding tasks, with deeper processing being associated with greater ERP amplitude. The greatest amplitude of a late positive ERP component occurred during correct recognition, regardless of type of encoding. Prefrontal cortical activation during encoding and retrieval has also been investigated by PET scanning (Tulving et al., 1994; Kapur et al., 1994). Tulving et al. (1994) found that deeper encoding operations were associated with activity in the left prefrontal cortex, while retrieval was associated with activity in the right prefrontal cortex.
The present study is concerned with possible cardiovascular indices of depth of processing. A few previous studies have yielded promising results. For example, Cohen and Waters (1985) compared a phonetic task to high- and low-level semantic tasks. The high semantic condition (i.e. deeper processing) yielded highest recall, followed by the low semantic and phonetic conditions. In addition, heart rate (HR) and skin conductance were found to differentiate among processing levels; HR and skin conductance were highest for the high semantic level, moderate for the low semantic task, and lowest for the phonetic task. The differences among processing levels increased during recall. Also, greater HR and skin conductance increases occurred on encoding trials which were later recalled.
Cacioppo et al. (1985) extended previous work by comparing four types of encoding tasks increasing in
processing depth: orthography, grammar, evaluation and self-referent. Self-report measures demonstrated that the grammar task required the most effort while the self-referent task required slightly less effort and the orthography task was rated as requiring low effort. Behaviorally, tasks rated as more effortful were associated with increased encoding latencies while recall was associated with processing level. HR and T-wave amplitude (TWA) during the recall phase of the memory task provided unexpected findings; the authors found no relationship between initial processing level and HR, but reported that deeper levels of processing were monotonically related to decreases in TWA. No cardiovascular changes were found during encoding. However, an experiment by Vincent et al. (1993) failed to find a relation between processing level and TWA, although they used a somewhat different procedure from that used by Cacioppo et al. (1985). In contrast, Vincent et al. (1993) found that increases in HR and decreases in TWA were associated with the difficulty of the memory task at encoding, and not with processing level. However, during recall no differences were uncovered in HR and TWA. Mulder and Mulder (1981) found that decreases in HR variability related to the amount of 'controlled processing' required during a cognitive task. Possibly some of the discrepancies seen among earlier studies may be due to differences in controlled processing induced by the different encoding and recall conditions.
The results from studies using cardiovascular measures are interesting, if somewhat inconsistent. One possibility is that these cardiovascular indices measure task difficulty or degree of perceived effort, possibly associated with changes in the amount of controlled processing, rather than qualitatively distinct cognitive processes. The relations among encoding task difficulty, depth of processing, and subsequent memory performance were examined by Craik and Tulving (1975). They contrasted an easy encoding task that necessitated deep (semantic) processing with a more difficult but shallower task, and found that memory was superior for the former task. That is, memory reflected depth of processing, not task difficulty. Using a similar contrast, Vincent et al. (1993) found that increases in HR and decreases in TWA were associated with task difficulty at encoding, rather than with depth of processing. How
ever, their design was unsatisfactory in that only one deep-easy task was contrasted with another shallow-difficult task. It would have been preferable to cross processing depth with task difficulty to evaluate the independent and interactive effects between effort and processing level.
The present experiments repair this omission. In the first experiment, three levels of processing were crossed with three levels of encoding task difficulty, and these encoding manipulations were followed by recognition memory tasks. Cardiovascular measures (blood pressure [BP], HR, HR variability, and TWA) were recorded during encoding and recognition, with the primary intention of clarifying the relations between the cardiovascular indices on the one hand and task difficulty, depth of processing, and memory performance on the other.
Task difficulty was manipulated by means of presentation rate (i.e., time pressure). The words to be learned for the later recognition test were presented at rates of 0.75 s, 1.5 s, or 3.0 s per word. The effects of these differences in workload 1 were measured both by means of a workload questionnaire which incorporated a scale of perceived effort and by performance on the recognition memory test. It was expected that faster presentation rates would be perceived as more demanding and would also be associated with lower memory performance.
It has been shown that the heart and vasculature show momentary adjustments throughout the processing of most perceptual and cognitive tasks (Jennings et al., 1990). The synchrony of these cardiovascular adjustments with task demands suggests that cardiovascular responses are induced by the requirements of information processing. The relations among HR, BP, and effort are reasonably well understood. Increases in mental effort are associated with increases in HR and BP and with decreases in TWA (Furedy, 1987; see Vincent and Furedy, 1992 for electrodermal effects). When sympathetic nervous system activity is increased pharmacologically, TWA is attenuated (Furedy and Heslegrave, 19831, and this attenuation may be reversed by beta-adren-
1 Workload is defined as the amount of processing per unit time.
ergic blockade (Rau, 1991). Therefore, an increase in beta-sympathetic activity leads to decreased TWA. Heart rate, being a mixed measure, controlled by both sympathetic and parasympathetic effects, is reactive to both sets of influences 2. HR variability suppression, is associated with a decrease in vagal tone which is predominantly parasympathetically mediated (Cacioppo et al., 1994; Porges and Byrne, 1992). The expectation in the present studies was that these cardiovascular measures would allow further differentiation of task difficulty, effort, and processing depth, than would be possible by the application of performance and self-report measures alone.
In summary, the purpose of the present experiments was to document the relations between several cardiovascular indices sensitive to task difficulty, the processes of memory encoding (i.e. level of processing), and memory performance. Specifically, we wished to examine the claims that deeper levels of processing were associated with changes in HR (Cohen and Waters, 19851, with decreases in TWA (Cacioppo et al., 1985), or perhaps with decreases in HR variability (Mulder and Mulder, 1981). Altematively, is it the case that cardiovascular indices are sensitive to increases in task difficulty and perceived effort, but are not sensitive to qualitative differences in encoding and recognition processes?
In addition, we hoped that the experiments would shed light on one issue in current memory theory. It has been suggested (e.g. Craik, 1983) that retrieval processes essentially recapitulate encoding processes; that is, an approximation to the encoding
2 We are aware that from a purely physiological perspective, the distinction between the sympathetic and parasympathetic branches of the autonomic nervous system disregards potential interactions between the systems. However, from a psychophysiological perspective, according to which the primary interest is in using physiological changes to study psychological processes (Furedy, 1983), the distinction between sympathetic and parasympathetic is heuristically useful. For example, the fact that TWA is attenuated by an iterative subtraction task, but not during the listening period when the numbers to be used are presented, whereas HR accelerates both during the listening period and during the task itself (Heslegrave and Furedy, 1979) indicates that the psychological processes of encoding and task completion may be distinguished in terms of the latter process involving sympathetic activity
event is reconstructed in the perceptual/cognitive system during recall (Lockhart et al., 1976). If this is the case, and different encoding tasks lead to different patterns of cardiovascular responding, these different patterns should also emerge during recognition.
These goals were met in Experiment 1 by crossing three levels of processing with three levels of task difficulty, accomplished by presenting the words to be learned at different rates. The differences in perceived effort associated with task difficulty were assessed by means of a workload questionnaire. Memory was assessed by means of a recognition test, and the cardiovascular indices of BP, HR, HR variability, and TWA were recorded during both encoding and recognition.
2. Experiment 1
In overview, participants made judgements about groups of 20 words in each of nine experimental conditions — three levels of processing crossed with three rates of presentation. These judgements, or orienting tasks, represented the encoding phase of memory. The recognition tests were presented following each experimental condition. The difficulty of each encoding condition was assessed immediately after the condition by means of a workload questionnaire. Finally, various cardiovascular measures were recorded during each encoding and recognition phase.
2.1. Method
2.1.1. Subjects
The subjects consisted of 31 male 3 university students, recruited with posted signs from the University of Toronto community. Subjects were informed that they would receive $8.00 for their participation in the experiment and that they were participating in a study of the physiological correlates of memory.
3 Male subjects were used to minimize intersubject variability. The psychophysiological results may not generalize to females.
2.1.2. Experimental task
In each experimental condition, subjects were first shown a question; they then answered that same question with respect to each of 20 words. The questions were answered either 'yes' or 'no' by pressing the relevant key from two response keys. The three question types used in the experiment operationally represented three levels of processing (Craik and Tulving, 1975); from shallow to deep the questions were 'Is the word in UPPER case?' 'Does the word contain one syllable?' and 'Is the word pleasant?' Approximately half of the words presented were in upper case and half were in lower case; half contained one syllable and half contained two syllables. In order to control for question effects, approximately half of the subjects were given the questions just described and half were given the complementary forms of the question; that is, 'Is the word in lower case?' 'Does the word contain two syllables?' and 'Is the word unpleasant?' Which specific form of the question appeared was randomly determined for each subject. Each of the three question types (case, syllable, pleasantness) was presented in three presentation rate conditions — words were presented for either 750 ms, 1500 ms, or 3000 ms.