The present study was conducted on 50 adult male Sprague-Dawley rats (250-280 g). The animal experiment was approved by the Ethics Committee of Shiraz University of Medical Sciences, Shiraz, Iran (Agreement Licensee No: 90-5954). The study animals were divided into five groups each containing 10 rats. Group 1 received distilled water by gavage. Group 2 received daily gavages of olive oil. Group 3 received daily gavages of curcumin (100 mg/kg/day) solved in olive oil [23]. Group 4 received daily gavages of sulfite in the form of sodium metabisulfite (25 mg/kg/day) solved in distilled water [15]. Group 5 received sodium metabisulfite (25 mg/kg/day) and curcumin by gavages. All the animals received daily gavages for 8 weeks. It should be mentioned that they were housed in plastic cages under standard conditions. The doses of sodium metabisulfite were selected according to the previous studies. Human body is exposed to sulfite through the ingestion of the sulfites that are used as preservatives in foods and beverages. Earlier studies have determined the acceptable daily intake from foods and beverages in a single day or meal (163 mg/day) by World Health Organization (WHO, 1994) [26-28]. The acceptable daily intake of 0-0.7 mg/kg was assigned to sulfur dioxide as well as sulfur dioxide equivalents arising from Na₂S₂O₅. It should be noted that the mean per capita of sulfite intake limit from foods and beverages is estimated as 19 mg sulfur dioxide equivalents per day. This level has been reported to be 163 mg in the 99^th percentile of the population [26-28]. The dose of curcumin was selected according to our earlier results. Our previous study showed 100 mg/kg/day as the appropriate dose of curcumin with no side effects on the liver, kidney, blood levels of aspartate aminotransferase, alanine aminotransferase, urea nitrogen, and creatinine [22, 29].

Working memory can be defined as a memory for an object, stimulus, or location that is used within a testing session, but not typically between the sessions. It is differed from reference memory which is a memory that would typically be acquired with repeated training and would persist from days to months. Reference memory is often the memory for the 'rules' of a given task. For example, a bar press gives a food pellet or a water maze contains a hidden platform or entrances into the baited arms of the radial maze. On the other hand, working memory allows the animal to remember which arms it had visited in a session. On the next day, this memory is no longer useful since the entire maze arms are baited again [30-32].

The Radial Arm Maze (RAM) was designed to measure the learning and memory in rats. The original apparatus consists of eight equidistantly spaced arms (42×12×12 cm³). The arms were designed to radiate from a central octagonal platform like the spokes of a wheel. Behavior experiments were performed by a blinded observer. The animals were assessed for learning and memory in a partially baited RAM. Prior to the training, the animals were kept on a restricted regime; so that their body weight would reach 85% of that prior to the training [30-32].

The animals were offered two sessions of adaptation on two continuous days prior to the beginning of the learning career. During these adaptation sessions, they were allowed to explore the baited arms of the maze for 10 min. After the adaptation sessions, the acquisition session was started.

During the acquisition session (learning career), the animals were given two acquisition trials per day until they attained the learning criteria. The learning criteria were defined as follows. The trial was continued for 5 min and the training was continued until the rats reached the criteria of 80% correct choice; i.e., at least four correct entries out of five. This session lasted for eight to fifteen days. At the beginning of each trial, the maze was cleaned with ethanol (70%) and thereafter four of the arms (2, 3, 5, and 7) were baited with food reward. The rat was placed on the central platform and was allowed a free action. When a rat ate bait or reached the end of an arm, the arm choice was recorded. Only the first accession to the baited arm was recorded as a correct choice and the maze arms were not rebaited. Entrances into the unbaited arms were recorded as Reference Memory Errors (RME), while reentrances into the baited arms were recorded as Working Memory Errors (WME). Each rat was given two trials daily and the data obtained from the two trials were averaged and entered into the final data analysis. The rats' performance was scored by the percentage of the correct choices, RME, and WME [30-32].

Ten days after acquisition, the rats were evaluated for retention of the task. The rats were given two trials and the mean scores of the percentage of the correct choices, WME, and RME was used for analysis [30-32].

The data is expressed as mean±SD. Either two-way or one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test was used to compare the means. Besides, p≤0.05 was considered as statistically significant.

In comparison to the controls, the performances of sodium metabisulfite-treated rats were impaired in acquisition of the task. Also, it took significantly more number of days for this group to acquire the criterion of 80% correct choices (p<0.001) (Fig. 1). Compared to the sodium metabisulfite treated rats, it took significantly fewer days for the rats receiving sodium metabisulfite and curcumin to reach the criterion of 80% correct choice (p<0.001) (Fig. 1). This suggests that curcumin prevents impairment in acquisition of the criterion of 80% correct choices induced by sodium metabisulfite.

A two-way repeated measures ANOVA was performed with training day (Day) as within-subjects factor and experimental group (Group) as between-subjects factor. A significant difference was found for Day [p<0.001), F (7, 315)=28.36], Group [p<0.001), F (4, 45)=24.42], and the Day by Group interaction [p<0.001), F (28, 315)=4.86]. This suggests that the rats' performance changed during the acquisition and that this change in performance was different among the groups (p<0.001) (Fig. 2). The rats in the sulfite-treated group showed less progress in the selection of the correct choices compared to the control animals from the 1^st to the 8^th day (Fig. 2). The score of the correct choices was increased in the rats of the sulfite+curcumin group in comparison to the sulfite-treated animals (Fig. 2).

Performance on Day 8 was further analyzed within the repeated measures ANOVA framework to assess the group differences in learning. The study results revealed a significant difference among different groups regarding the percentage of the correct choices at this time point [p<0.001), F (4, 45) =72.75], Furthermore, in comparison to the control groups, the sulfite group showed a significant reduction in the percentage of the correct choices (p<0.001). Nevertheless, no significant difference was found between the sulfite+curcumin group in comparison to the control groups (Table 1). This demonstrated that concomitant treatment of curcumin during sulfite consumption prevented the reduction of scores of the correct choices in the acquisition session.

A significant effect was observed for [p<0.001), F (7,315) =19.23], Group [p<0.001), F (4, 45) =4.45], and the Day by Group interaction [p<0.001), F (28,315) =4.17]. This suggests that the RME during the acquisition phase changed with the days of training and these changes were different among the study groups (Fig. 3). The rats in the sulfite-treated group showed more reference memory errors compared to the control animals from the 1^st to the 8^th day (Fig. 3). Fewer errors were observed in evaluation of the reference memory in the rats of sulfite+curcumin group in comparison to the sulfite-treated animals (Fig. 3).

RME on the 8^th day of the acquisition session were analyzed to assess the group differences in learning. A significant difference was found among different groups regarding the number of reference memory errors at this time point [p<0.001), F (4, 45) =51.85]. The study results revealed significantly more reference memory errors in the sulfite group on day 8 compared to the control groups (p<0.001). Nonetheless, no significant difference was found between the sulfite+curcumin group and the control groups (Table 1). This demonstrated that concomitant treatment of curcumin during sulfite consumption caused fewer errors during evaluation of the reference memory in the acquisition session.

A two-way repeated measures ANOVA was performed with the training day (Day) as within-subjects factor and the experimental group (Group) as between-subjects factor. A significant effect was observed for [p<0.001), F (7,315) =18.72], Group [p<0.001), F (4, 45) =33.86], and the Day by Group interaction [p<0.001), F (28,315) =5.71] (Fig. 4). The rats in the sulfite-treated group showed more working memory errors compared to the control animals from the 1^st to the 8^th day (Fig. 4). Fewer errors were observed in evaluation of the working memory in the rats of sulfite+curcumin group in comparison to the sulfite-treated animals (Fig. 4).

WME on day 8 were analyzed to assess the group differences in learning. A significant difference was found among different groups regarding the number of WME [p<0.001), F (4, 45) =19.81]. Moreover, the sulfite group showed significantly more WME on the 8^th day in comparison to the other groups (p<0.001). No significant difference was found between the sulfite+curcumin group and the control groups (Table 1). This demonstrated that concomitant treatment of curcumin during sulfite consumption caused fewer errors during evaluation of the working memory in the acquisition session.

A one-way ANOVA was used to analyze the percentage of the correct choices during the retention session. A significant difference was found among the study groups regarding the percentage of correct choices [p<0.001), F (4, 45) =37.80] (Fig. 5). Furthermore, the sulfite group revealed a significant reduction in the percentage of the correct choices compared to the control groups (p<0.001). Nevertheless, no significant difference was found between the sulfite+curcumin group and the control groups (Fig. 5). This demonstrated that concomitant treatment of curcumin during sulfite consumption prevented the reduction of scores of the correct choices in the retention session.

A significant difference was observed among the groups regarding the number of reference memory errors [p<0.001), F (4, 45) =38.00] and the number of working memory errors [p<0.001), F (4,45) =44.85]. Besides, in comparison to the other groups, the sulfite group showed more RME and WME during the retention testing (p<0.001) (Fig. 6, 7). No significant difference was found between the sulfite+curcumin group and the control groups. This demonstrated that concomitant treatment of curcumin during sulfite consumption caused fewer errors during evaluation of the reference and working memories in the retention session (Fig. 6, 7).