The prefrontal cortex (PFC) has a central role in working memory (WM). Resistance to distraction is considered a fundamental feature of WM and PFC neuronal activity. However, although unexpected stimuli often disrupt our work, little is known about the underlying neuronal mechanisms involved. In the present study, we investigated whether irregularly presented distracters disrupt WM task performance and underlying neuronal activity. We recorded single neuron activity in the PFC of 2 monkeys performing WM tasks and investigated effects of auditory and visual distracters on WM performance and neuronal activity. Distracters impaired memory task performance and affected PFC neuronal activity. Distraction that was of the same sensory modality as the memorandum was more likely to impair WM performance and interfere with memory-related neuronal activity than information that was of a different sensory modality. The study also shows that neurons not involved in memory processing in less demanding conditions may become engaged in WM processing in more demanding conditions. The study demonstrates that WM performance and underlying neuronal activity are vulnerable to irregular distracters and suggests that the PFC has mechanisms that help to compensate for disruptive effects of external distracters.
Working memory (WM) refers to the ability to maintain in “memory” for a short period of time information that is relevant to the guidance of behavior (Baddeley 1986). Lesion (Jacobsen 1936; Mishkin 1957; Goldman and Rosvold 1970; Passingham 1975) and electrophysiological studies in nonhuman primates (Goldman-Rakic 1987; Fuster 1997) have indicated that the prefrontal cortex (PFC) has a central role in the performance of tasks that require short-term memory. Single PFC neurons respond to mnemonic processing of the location or content of visual (Fuster 1973; Niki 1974; Funahashi et al. 1989; Wilson et al. 1993; Rao et al. 1997) and auditory stimuli (Bodner et al. 1996; Kikuchi-Yorioka and Sawaguchi 2000; Artchakov et al. 2007) and form cross-modal associations between tones and colors (Fuster et al. 2000).
In natural circumstances, we seldom work in complete quiescence. Distracting stimuli interfere with goal-directed behavior and impair WM performance (Anourova et al. 1999; Vuontela et al. 1999). The level of performance that can be achieved in the presence of interfering stimuli depends on the ability to maintain task relevant information in mind and, on the other hand, on the ability to keep the distracters from interfering with the ongoing goal-directed processes (Bunge et al. 2001).
The PFC has been suggested to have a critical role in the maintenance of information and in the resistance of the effects of intervening stimuli on goal-directed behavior. Evidence supporting the role of the frontal areas in the control of distraction comes from lesion and electrophysiological studies performed in monkeys (Malmo 1942; Miller et al. 1996) and in humans (Chao and Knight 1995, 1998, Chiu et al. 2008) and from recent neuroimaging studies in humans (Sakai et al. 2002; Dolcos et al. 2006). Monkeys with bilateral lesions in the PFC were able to perform visual short-term memory tasks in darkness but not in daylight conditions, which was interpreted to indicate inability of the lesioned monkeys to resist distraction (Malmo 1942). Only a few electrophysiological studies have investigated the effects of distracters on single neuron activity in the monkey PFC. In one early study describing neuronal activity in PFC, variable distracters were presented during the performance of a delayed response task (Fuster 1973). The distracters were described to impair performance and attenuate delay-related activity in the PFC neurons. Another study, on the other hand, using electrophysiological recordings in monkeys performing delayed matching to sample (DMTS) memory tasks provided different kinds of results by showing that PFC neurons maintained or increased their firing rate during the memory maintenance period despite intervening stimuli between the sample and the target stimulus (Miller et al. 1996). In that study, monkeys responded to the first match stimulus but not to the intervening nonmatch stimuli. The task can be seen to have mimicked an everyday situation of working in noisy environments. However, WM-related neuronal activity may be less resistant to distraction in a situation in which the performance is disrupted by irregularly presented distracters.
Recent human brain imaging studies have shown that distracters that are presented during the memory maintenance period of a WM task may have variable effects on the delay activity in the dorsolateral PFC. Emotional distraction (Dolcos and McCarthy 2006) or distracters that were nonconfusable with the memoranda (Dolcos et al. 2008) decreased the delay-period activity, whereas distracters that could be confused with the memoranda increased the delay-period activity in the dorsolateral PFC (Dolcos et al. 2008). Repeated presentation of equally familiar stimuli was shown to enhance or reduce cortical activity depending on the role of the stimuli in WM—whether memoranda or distracters (Jiang et al. 2000).
In the present study, we investigated the effects of randomly presented distracters on the performance of WM tasks and the underlying neuronal activity. Single neuron activity was recorded in the dorsolateral PFC of 2 monkeys trained to perform DMTS tasks. This area was chosen because our principal aim was to investigate how randomly presented distracters affect memory-related neuronal activity, and many earlier studies have demonstrated that the area in and around the principal sulcus is involved in short-term memory processing (e.g., Fuster and Alexander 1971; Fuster 1973; Niki 1974; Goldman-Rakic 1987; Funahashi et al. 1989; Carlson et al. 1990, 1997; Artchakov et al. 2007). We hypothesized that task-irrelevant stimuli presented randomly during the memory maintenance period would affect the memory-related neuronal activity and impair memory task performance. We also expected that distracters that are of the same sensory modality as the memorandum would affect memory-related neuronal activity more than distracters of a different sensory modality.
Materials and Methods
The recording methods and the surgical procedures for implantation of the cylinder have been described in detail earlier (Artchakov et al. 2007) and will be described in short here. Extracellular single cell activity was recorded with tungsten microelectrodes (impedances 10–20 MΩ, Frederick Haer & Co., Bowdoinham, ME) in the right dorsolateral PFC in 2 female rhesus monkeys (Macaca mulatta, 5.5 kg and 4.0 kg) trained to perform visual (2 monkeys) and auditory (1 monkey) spatial DMTS tasks (Fig. 1a). For the recordings, a cylinder (diameter 18 mm) was implanted over the PFC covering the cortical area of and around the principal sulcus (center coordinates: anterior 30–31, lateral 14–15) (Fig. 1b). During the recording session, the monkey's head was fixed to the chair with a halo and its left arm was restrained. The presentation of the behavioral task and the collection of the data were controlled by a software package (Cortex 3.1). Visual fixation of the central light emitting diode (LED) was required throughout the trial until the reward (a drop of juice) was delivered. The visual fixation was considered broken when the monkey's gaze moved outside of a central eye window (programmed to the size of about 2–4°), and the trial was automatically discontinued. To confirm the location of the recorded neurons in the PFC, a high-resolution structural magnetic resonance image set using a 1.5 T scanner (Siemens Sonata) was obtained from both monkeys (Artchakov et al. 2007).
The firing of a single neuron was isolated from the amplified and filtered neuronal activity and digitized by a window discriminator. The data were analyzed off-line, and rasters and peristimulus time histograms were produced for each condition by aligning the data at the following temporal events: start of visual fixation, start of lever pulling, sample cue onset, and distracter onset.
The animals were housed in individual cages in the same room with the other monkeys of the colony. During training and recording, the monkeys were water deprived in their home cages and rewarded with juice during task performance. The animals were weighed daily during the training and recording periods. Their fluid intake was closely monitored and if they did not perform the task until satiety, additional water was given. The animals also received fruits or vegetables everyday.
The DMTS Tasks
A trial started when the monkey pulled a lever with the right hand and visually fixated a central LED on a screen (59 × 35 cm) 55 cm from the animal. Eye movements were recorded with an infrared eye movement tracker (Microguide Inc., Downers Grove, IL). Visual fixation of the central LED was required throughout the trial until the reward (a drop of juice) was delivered. After 700 ms of visual fixation and holding the lever, a visual (in the visuospatial DMTS task) or an auditory (in the audiospatial DMTS task) sample cue (duration 500 ms) was presented either on the “left” or on the “right” side of the central LED at an eccentricity of 10°. In visual trials, the cue was a green LED, and in auditory trials, a tone (500 Hz, mean loudness ca. 72 dB SPL, measured with Precision Sound Level Meter Type 2203; Bruel & Kjaer, Naerum, Denmark) from 1 of 2 loudspeakers located on the left and right side of the monkey. Both monkeys performed the visuospatial DMTS task, and one of the monkeys also performed the auditory DMTS task (Fig. 1a). When the monkey performed both the visual and auditory versions of the DMTS task, the auditory and visual trials were presented in a random order. During the delay period of the DMTS tasks (duration 3100 ms), the monkey continued holding the lever and maintaining in memory the location (left or right) of the sample cue. A probe stimulus (duration 300 ms) of the same modality as the sample cue was presented at the end of the delay. When the probe stimulus appeared at the same location as the sample cue (match trials), the monkey was to release the lever within 1000 ms to get a reward, and if the location of the probe stimulus was different from that of the sample cue (nonmatch trials), the monkey was to continue holding the lever for ≥1000 ms after the probe stimulus to get the reward. In randomly presented trials, a visual or an auditory distraction (duration 100 or 300 ms) was presented in the middle of the delay period. In a subgroup of neurons, a distracter was presented at the beginning (100 ms from the offset of the sample cue) or in the middle of the delay period. These 2 kinds of distracted trials as well as control trials were presented in a random order. The visual distraction was produced by the simultaneous lighting of 8 LEDs; these LEDs were evenly distributed around the central LED at an eccentricity of 10°. Two of the 8 LEDs were also used as the left and the right cues in the visual WM trials. The auditory distraction was composed of a sound (the tone that was used in the auditory trials) that was presented intermittently from the left and the right loudspeakers for 10 ms (alternating between the left and right loudspeakers with a frequency of 50 Hz).
Analysis of Behavioral Data
Accuracy and reaction times (RT) were analyzed statistically using a 1-way analysis of variance (ANOVA) followed by Tukey–Kramer post hoc test when appropriate. P < 0.05 was considered a statistically significant result. Behavioral results are based on the performance during recording sessions. All recorded sessions were used in the analysis.
Analysis of Neuronal Data
Analyses of neuronal activity were performed for epochs containing the baseline activity (the 700 ms period before the presentation of the sample cue during which the monkey was visually fixating and holding the lever), sample cue (duration 500 ms), and delay period (3100 ms from the offset of the sample cue). Left match and nonmatch trials (and the corresponding trials in which the memorandum was on the right) were combined because the epochs covering the presentation of the probe stimulus and motor responses were excluded from the present analysis. For statistical analyses, the delay period was divided into shorter time intervals in order to capture responses to the distracter that in some neurons were relatively short lasting. When the distracter was in the middle of the delay, the delay was divided into 7 time epochs: three 466-ms time intervals before and 3 equally long time intervals after the distracter and the distracter period. In the subgroup of neurons in which the distracter was presented at the beginning of the delay, the delay after the distracter was divided into 6 equally long intervals (450 ms) starting from the offset of the distracter. Neuronal activity was analyzed with a custom made script on Matlab 7.3.0 (The MathWorks Inc., Natick, MA).
We used a 1-way ANOVA with Tukey–Kramer post hoc test, the recommended post hoc test for 1-way ANOVA in GraphPad Instat (GraphPad Instat, 3.00, 1997; GraphPad Software Inc., La Jolla, CA) and Matlab (Matlab, version 18.104.22.1687; The MathWorks Inc.), to examine whether neuronal activity in the control trials during the sample cue and delay period differed significantly from the baseline activity. In all statistical tests, P < 0.05 was considered statistically significant. When the discharge rate during one or more time intervals of the task differed significantly from the baseline activity, the neuron was considered task related. If the firing rate during the sample cue in the control trials of the task-related neurons differed significantly from the baseline activity, the neuron was classified as cue related. If the firing rate during one or several intervals of the delay period differed significantly from the baseline, the neuron was classified as delay related. To find out whether cue- and delay-related neurons were spatially tuned, we performed 2-way ANOVA with factors interval (8 intervals: the sample cue interval and 7 intervals of delay) and location (left and right) followed by Dunn–Sidak post hoc test when appropriate. If there was a significant difference between the responses to the left and right sample cues, the neuron was considered to have a spatially tuned cue response. If there was a significant difference between the neuronal activity during the left and right delay, the neuron was considered to have spatially tuned delay-period activity. Finally, to find out whether distraction affected the neuronal firing rate, we performed 3-way ANOVA with factors interval (distracter and 3 or 6 intervals of delay after the distracter in the middle or at the beginning, respectively), location, and distracter (no distracter, visual distracter, auditory distracter) followed by Dunn–Sidak post hoc test when appropriate. Dunn–Sidak was used with 2- and 3-way ANOVA. It is apparently less conservative than the Bonferroni test that is not recommended when more than 5 comparisons are performed (GraphPad Instat, 3.00, 1997; GraphPad Software Inc.) as in 2- and 3-way ANOVA in the present study. When a neuron displayed delay-related (spatially tuned or nontuned) activity in control but not in distracted trials or when the spatial tuning in distracted trials was significantly reduced compared with control trials, the activity was considered impaired by the distraction. When a neuron displayed spatially tuned delay-related activity in distracted but not in control trials, the spatial tuning was considered “created” by the distraction.
To calculate the onset latency of the responses to the sample cue, the spiking activity of the neurons was low-pass filtered by convolution with Gaussian distribution with σ of 20 and resampling resolution at 1 ms (Artchakov et al. 2007). A spike density function was created for the periods of the baseline and the sample cue (MacPherson and Aldridge 1979; Wiener and Richmond 1999), and a confidence interval (CI) of 95% from the baseline was used as a threshold for departure from the baseline. Low-pass filtering, spike density functions, and CI were generated with a custom made script on Matlab 7.3.0 (The MathWorks Inc.). The onset latency was determined as the time point at which the value of the function was 50% of the height from the 95% CI threshold line to the peak of the response (MacPherson and Aldridge 1979). The onset latency of the responses to the distracters was calculated in a similar manner except that the responses recorded during the control trials were first subtracted from the responses to the distracters. The duration of the responses was determined as the time between the onset and offset latencies (the time point at which the value of the function returned—after the peak value—to the 50% of the height from the 95% CI threshold line).
The maintenance of the monkeys and all procedures of the study were carried out according to the Finnish law and statutes governing animal experimentation. The Finnish Ministry of Agriculture had approved of the study and granted permission to perform it.
The monkeys performed the DMTS tasks without distraction with a mean accuracy of over 90% correct responses. Distracters impaired the task performance by increasing the RT and the number of incorrect responses (Fig. 2a) and could result in a failure to respond. Distracters disrupted the performance of the DMTS task more when the sensory modality of the distracter was the same as that of the memorandum. Thus, visual distracters impaired the performance of the visual task significantly more than auditory distracters and auditory distracters impaired the performance of the auditory task more than visual distracters.
When the visual distracter occurred randomly either at the beginning or in the middle of the delay in the visual task, the accuracy was similar in both conditions but significantly poorer than in the control trials and in the trials with auditory distracters. When the visual distracter was in the middle of the delay, the RT was significantly longer than in all other conditions (Fig. 2b).
A total of 245 neurons were recorded during the visuospatial DMTS task and 67 of them also during the auditory version of the task. The classification of the neurons to different categories, based on their activity during the control trials, is presented in Table 1. The mean onset latency of the responses to the visual sample cue was 151 ± 98 ms (mean ± standard deviation [SD]) and to the auditory sample cue was 153 ± 110 ms.
|Type of response||Number of neurons||% of neurons|
|Cue, spatially selectivea||7||3|
|Cue, spatially nonselectivea||0||0|
|Delay, spatially selective||36||15|
|Delay only, n = 18 (7%)|
|Cue + delay, n = 18 (7%)|
|Delay, spatially nonselective||27||11|
|Delay only, n = 22 (9%)|
|Cue + delay, n = 5 (2%)|
|Not task related||175||71|
|Type of response||Number of neurons||% of neurons|
|Cue, spatially selectivea||7||3|
|Cue, spatially nonselectivea||0||0|
|Delay, spatially selective||36||15|
|Delay only, n = 18 (7%)|
|Cue + delay, n = 18 (7%)|
|Delay, spatially nonselective||27||11|
|Delay only, n = 22 (9%)|
|Cue + delay, n = 5 (2%)|
|Not task related||175||71|
Neurons responding during the cue but not during the delay.
Figure 1c illustrates the proportions of the neurons that were affected by distraction. Distracters altered neuronal activity statistically significantly in 97 neurons (40% of the recorded neurons). Of them, 51 neurons were task related in the control trials (73% of the task-related neurons). Particularly, the memory neurons, that is, neurons that exhibited spatially tuned activity during the delay period, reacted to the distracters (81%, n = 29 of the 36 neurons exhibiting spatially selective activity during the delay in control trials). In comparison, only 26% of the neurons that showed no DMTS task-related activity (n = 46 of 175 neurons) were affected by distracters.
Distracters Impaired Delay-Related Neuronal Activity
In almost half of the neurons exhibiting, in the control condition, spatially tuned activity during the delay period (44%, n = 16 of 36 neurons), the spatial tuning was impaired in the distracted trials. In most of these neurons (94%, n = 15 of 16 neurons), only a distracter that was of the same sensory modality as the sample cue (visual distracter in visual task and auditory in auditory task) impaired the delay-related spatial tuning (Fig. 3). In 4 delay-related neurons that were not spatially tuned in the control trials but that changed their activity significantly during the delay in both left and right trials (15% of the 27 neurons), distracters impaired the delay activity so that it no more differed significantly from baseline activity.
Distracters Created Spatially Tuned Delay-Related Neuronal Activity
Some neurons (n = 7) that displayed no spatial tuning during the delay period in control trials developed spatial tuning during the delay as a result of the distraction (5 neurons by visual and 1 by auditory distraction in the visual task and 1 by auditory distraction in the auditory task). These neurons changed their spiking activity significantly during the delay period after the distracter in trials where the monkey memorized one location but not in trials where the monkey memorized the other location (Fig. 4). In control trials, 5 of the 7 neurons exhibited no task-related activity, 1 responded spatially selectively during the sample cue period, and 1 displayed delay-related activity that was not spatially tuned.
Other Neuronal Responses to the Distracter
A sizable number of neurons responded to the distracter but did not change the spatial tuning of their delay-period activity (n = 70, 72% of the 97 neurons that were affected by distraction). These neurons responded to the distracter similarly in all trials regardless of the spatial location of the sample cue. In the control trials, more than half of these neurons were not task related (n = 41, 59% of the 70 neurons), 15 neurons (21% of the 70 neurons) responded to the sample cue, and 11 of them also during the delay period. Another 14 neurons (20% of the 70 neurons) were delay related in the control trials. Of the responses to the distracter, 47 were to the visual, 17 to the auditory, and 6 to both kinds of (auditory + visual) distracters (Fig. 5). The mean onset latency of the responses to the visual distracter was 232 ± 253 ms (mean ± SD) and to the auditory distracter was 168 ± 227 ms. The mean onset latencies of the responses to the visual and auditory distracters were not statistically significantly different from each other (t-test, P > 0.05). The mean duration of the responses to the visual distracter was 384 ± 377 ms and to the auditory distracter was 482 ± 493 ms. In 25% of the neurons responding to the visual distracter and in 33% to the auditory distracter, the response to the distracter was long lasting (>600 ms), whereas in the remaining neurons it was <400 ms. Most neurons responded to the distracter by increasing their activity but in 10 neurons the response to the visual and in 5 neurons to the auditory distracter was inhibitory.
Effect of Distracter Timing on the Neuronal Responses
In a subgroup of neurons (n = 59), the distracter was presented randomly at the beginning (at 100 ms from the offset of the sample cue) or in the middle of the delay. This subset of neurons was recorded in one monkey performing the visual DMTS task. Distracters altered the neuronal activity in 54% of these neurons (n = 32 of 59 neurons). In 18 neurons, the timing of the distraction was crucial: In 13 neurons, a distracter at the beginning but not in the middle of the delay significantly affected the neuronal firing and in 5 neurons in the middle but not at the beginning of the delay.
In 6 neurons, the visual distracter impaired spatial tuning of the delay-period activity. The distracter impaired the activity when presented at the beginning in 4 neurons, when presented in the middle in 1 neuron and when presented either at the beginning or in the middle in the other neuron. Two neurons exhibited a creation of spatially tuned delay activity when the distracter was presented at the beginning but not in the middle of the delay (Fig. 4). In both these neurons, only the visual distracter was effective.
Neuronal Baseline Activity
The distribution of the mean baseline activity of the recorded neurons is illustrated in Figure 6a. Most neurons had a mean baseline activity below 10 Hz. The mean baseline activity was significantly higher in those neurons in which the distracter either impaired or created spatial tuning or produced other types of neuronal responses than in neurons that were not affected by distracters (Fig. 6b).
Several lines of evidence have suggested that one of the functions of the PFC is to protect memory from adverse effects of distraction. Lesions in the monkey and human PFC render subjects vulnerable to distraction and impair their performance in tasks requiring short-term memory (Malmo 1942; D'Esposito and Postle 1999). Single cell recordings in monkeys performing WM tasks have shown that sustained activity in the PFC neurons is maintained despite intervening stimuli (Miller et al. 1996). The results of the present study add knowledge about the effects of distraction by demonstrating that randomly presented irrelevant stimuli occurring during the WM task impair behavioral performance and interfere with the underlying neuronal activity. Furthermore, some neurons in the PFC create spatially tuned neuronal activity as a result of distraction suggesting recruitment of neurons to memory processing when the task becomes more demanding.
In the present study, several neurons in the control condition responded spatially selectively during the cue and/or delay period. This finding is in line with earlier literature on electrophysiological recordings in monkeys suggesting a role for the PFC in WM processes and temporal integration of information (Fuster 1973, 1997; Goldman-Rakic 1987; Funahashi et al. 1989; Fuster et al. 2000). The proportion of neurons that, in the control condition, responded spatially selectively during the delay period (7%) or during both cue and delay periods (7%; Table 1) is comparable to earlier electrophysiological studies in monkeys in which neuronal activity was recorded in the PFC during a spatial delayed response or DMTS task with 2 locations (e.g., Niki 1974; Batuev et al. 1985; Artchakov et al. 2007). When more than 2 locations were used as memoranda the proportion of neurons exhibiting spatial selectivity during the delay has been higher (Funahashi et al. 1989). Also in accordance with earlier literature (Funahashi et al. 1989, 1990; Artchakov et al. 2007), most neurons responding during the cue period were spatially selective in the present study: all neurons responding only during the cue period and 78% of those responding both during the cue and the delay.
As in human studies (Anourova et al. 1999; Vuontela et al. 1999; Sakai et al. 2002; Dolcos and McCarthy 2006; Dolcos et al. 2007), distracters in the current study reduced the accuracy and increased the RTs of the memory task performance. The finding that visual distracters impaired the visual task performance of the monkeys more than auditory distracters and auditory distracters the performance of the auditory task more than the visual task corroborates results from human studies suggesting that information that is qualitatively similar to the memorandum impairs WM performance more than qualitatively different information (Anourova et al. 1999; Vuontela et al. 1999).
The reduced performance level in the DMTS task observed in the current study can be understood by the results of the single cell recordings in the PFC. Distracters impaired the memory-related delay-period activity in several neurons thus confirming our first hypothesis. This finding is in line with an observation by Fuster (1973) but appears contradictory to the result of an earlier study reporting that PFC neurons maintain or increase sample selective delay activity when intervening stimuli were presented (Miller et al. 1996). The differences between the results of these studies are likely due to differences in the experimental set up. In the current study, the control trials and trials with distracters were presented in a random order to the monkey, whereas in the study by Miller et al. (1996), a variable number of intervening stimuli occurred during each trial. Furthermore, in their study, the monkeys were trained to respond only to the match stimulus and ignore the intervening nonmatch stimuli, whereas in the current study, match and nonmatch trials were presented in a random order and, during the delay period, the monkey was unaware of the type of response that was required at the end of the trial.
In the current study, neurons that exhibited spatially tuned activity during the delay period in the control condition were particularly sensitive to the effects of distracters. In most of these neurons, as we had predicted, only a distracter that was of the same sensory modality as the sample cue (visual distracter in visual task and auditory in auditory task) impaired the spatial tuning (Fig. 3). This finding was also reflected in the task performance: distracters qualitatively similar to the memorandum impaired WM performance more than qualitatively different distracters. The result that cross-modal distracters disturbed neural activity less also indicates that auditory and visual information in the PFC is processed independently in neural networks involved in attention and memory processing, corroborating results of 2 earlier studies in monkeys (Kikuchi-Yorioka and Sawaguchi 2000; Artchakov et al. 2007).
In addition to its role in memory maintenance and executive functions, the PFC has been suggested to participate in an involuntary switch of attention produced, for example, by tones deviating from a repetitive sequence of standard tones (Näätänen 1990). This suggestion has been supported by studies on patients with PFC lesions demonstrating reduced attentional control of irrelevant sensory input (Chao and Knight 1995) and reduced amplitudes in responses to deviant tones in patients (Alho et al. 1994; Alain et al. 1998) and by brain imaging studies suggesting that in addition to a generator in the temporal lobes, a generator in the frontal lobes is involved in the automatic processing of acoustic changes (Opitz et al. 2002, Schönwiesner et al. 2007). The result of the present study that irregular distracters impaired the performance of WM tasks by increasing RTs and reducing performance accuracy may be partly explained by the proposed role of the PFC in change detection. Unexpected stimuli may be detected and evaluated in the PFC and—depending on the nature of the intervening stimulus—may or may not require switch of attention. These processes might reduce the capacity of the PFC to simultaneously engage in WM and goal-directed executive functions resulting in slowed RTs, reduced accuracy in WM task performance and impaired underlying neuronal activity.
Impairment of spatially tuned delay-period activity was not the only indication of the profound effects of distracters on PFC neuronal activity. Interestingly, in a subgroup of neurons that in the control condition showed no spatially selective delay-period activity, the distracter created spatially tuned delay-period activity. This finding suggests that, in the PFC, neurons that in less demanding conditions are not involved in memory processing may in more demanding situations become engaged in WM processing. In a recent neuroimaging study in humans, in which the performance of a spatial WM task was distracted by the performance of a second task, the activity in area 46 during the memory maintenance period was stronger on trials performed correctly than on those on which errors were made, which was suggested to indicate protection against distraction (Sakai et al. 2002). The present findings that some neurons developed spatial tuning and several other neurons maintained their spatial tuning during the delay period, when the task became more demanding as a consequence of distraction, reflect neural mechanisms providing resistance to distraction.
How did the distracter create spatially tuned activity? A possible mechanism is that neurons that in the control condition did not exhibit memory-related activity but created spatially tuned activity as a result of distraction were part of a network of neurons that was sensitive to the location of the memorandum exhibiting spatially tuned activity during the delay. Some neurons in this network may remain nonresponsive to the preferred location because the input activity carrying spatial information to these neurons is not strong enough to drive them to the threshold to fire action potentials. These neurons may receive inhibitory input from nearby neurons of the network to keep them from responding during a relatively simple memory task. When the task became more demanding (in the distracted trials), these neurons may have been disinhibited.
Several lines of evidence suggest that local cortical inhibitory circuitries play an important role in the construction of spatial tuning of cortical neurons (Eysel et al. 1990; Alloway and Burton 1991; Goldman-Rakic 1995; Sato et al. 1996; Murthy and Humphrey 1999). Furthermore, reorganization of cortical motor maps has been suggested to involve intracortical connections and critically placed inhibitory circuits (Jacobs and Donoghue 1991). In a study by Rao et al. (2000), it was demonstrated that γ-aminobutyric acid (GABA)A-mediated inhibition in the PFC has a central role in processes underlying WM. In their study, iontophoretic application of bicucullin methiodide, a GABAA receptor antagonist, to the PFC of monkeys destructed the normally existing spatial tuning or, in some cases, created spatial tuning in neurons that in control conditions displayed no directional selectivity. Injections of bicucullin to the PFC have also been shown to disrupt WM performance (Sawaguchi et al. 1988, 1989). The effects of the above described GABAA receptor antagonist are similar to the effects of distraction in our study (i.e. distracting stimuli impaired WM performance and impaired or created spatial tuning during the delay period in PFC neurons). This suggests that the effects of distraction may partly be explained by mechanisms that alter inhibitory synaptic activity. Moreover, the nature of the distracters in the current study was such that it may have further pushed these neurons to participate in the memory maintenance activity. The visual distracter was produced by the simultaneous lighting of 8 LEDs, 2 of which were also used as the left or right memorandum. The auditory distracter was composed of a sound presented intermittently from the left or right loudspeaker. The distracter, therefore, may have activated the network that was sensitive to the location of the memorandum causing disinhibition of some neurons that were part of this network but exhibited spontaneous activity in the control condition, increasing the probability that they will exhibit spatially selective delay activity. Because the visual distracter consisted of 8 LEDs (in addition to the left and right, it had upper, lower, and diagonal LED positions), some neurons that did not show any activity during the cue or delay periods in the control task, but responded to the distracter presentation (Fig. 5a), might have visual receptive fields in other than left or right visual fields (upper, lower, or diagonal locations). However, the neurons that created spatially selective delay activity after the presentation of the distracter most likely do not have visual fields in the upper, lower, or diagonal LED locations of the distracter. This is because in such case these neurons should have responded to the distracter in all visually distracted trials and not only in the left or right trials. It is the spatial selectivity of the created delay activity that suggests that these neurons were part of a neuronal network which was selective to the location (left or right) of the memorandum. This selectivity was not seen in the control task but was uncovered by the distracters.
Because there is some indication that the early part of the delay period in a memory task may be more sensitive than the later phases of the delay to manipulative interventions such as transcranial magnetic stimulation (Harris et al. 2002), we investigated the importance of the timing of distraction on a subgroup of PFC neurons when 1 monkey performed the visual DMTS task. The visual distracter at the beginning of the delay reduced the accuracy to the same extent as in the middle of the delay, but the distracter in the middle of the delay produced the longest RTs. However, a visual distracter at the beginning of the delay period affected the firing patterns of the neurons more often than in the middle of the delay. Thus, based on these results, it is not possible to judge whether distraction in one time point of the delay is more likely to modulate the memory-related neuronal activity than in any other time point, and further studies are needed to resolve the question.
Most recorded neurons in our study had a relatively low mean baseline firing rate (<10 Hz) (Fig. 6a). Grouping the neurons according to their responsiveness to the distracters showed a significant difference in the mean baseline frequency between these groups (Fig. 6b) suggesting that neurons with higher baseline activity were more likely to be affected by distraction. During the recording, however, we did not systematically classify the neurons to fast or regularly spiking based on the duration of the action potential, discharge rate, and spatial extent of the potential field (e.g., Mountcastle et al. 1969). Therefore, it remains to be determined whether, for example, fast spiking neurons are more likely to be responding to the distracters than regularly spiking neurons.
The monkeys were well-trained in the DMTS task (performance level without distracters was >90% correct responses), but their performance dropped statistically significantly, when distracters were presented. The short-lasting distracter, however, did not totally destroy the performance, and the monkeys were able to perform the task still relatively well (Fig. 2). In line with this behavioral result, it is noteworthy that distracters impaired the firing patterns of only a subset (44%) of the neurons that exhibited memory-related activity. Furthermore, as a result of distraction, some neurons became involved in memory processing exhibiting spatially selective delay-period activity. The activity of these neurons may have compensated for the adverse effects of the distracters. The overall effect of the distraction, however, at neuronal level and behaviorally, was impairment of WM. These results show a close relationship between the level of behavioral performance and the underlying neuronal activity. It is likely that a more robust distracter might have impaired the behavioral performance and underlying neuronal activity more than the short-lasting distracters of the present study.
A possible limitation of the study is that the recordings in both monkeys were performed in the right hemisphere. This was due to the history of the monkeys: In both monkeys, recordings had been performed earlier in the left PFC. Thus, regarding the neuronal responses, we can neither generalize the findings of the study to both hemispheres nor make any statement about the lateralization of the responses.
In conclusion, the present study demonstrates that distracters presented during the memory maintenance period of a WM task may profoundly affect memory-related neuronal activity in the PFC, probably leading to the impaired performance. Distraction that has the same sensory modality as the memorandum is more likely to impair WM performance and interfere with memory-related neuronal activity than information that arises from a different sensory modality. The study also shows that neurons that in less demanding conditions are not involved in memory processing may in more demanding conditions become engaged in WM processing, suggesting a compensatory neuronal mechanism within PFC for the adverse effects of distraction on WM.
Academy of Finland (National Centers of Excellence Program 2006–2011 and Chinese–Finnish Collaboration of the Neuro Program No. 214412); National Science Foundation of China (Chinese–Finnish International Collaboration, Project-neuro No. 30621130076, 30530270); University of Helsinki.
Conflict of Interest: None declared.