1 trial :  nothing is given for result comparision - so we have no idea if it's a mistake. 
2nd trial : The results can be compared - if varies, one may go wrong, but which one? 
3rd trial : If 3rd result is different from 1st and 2nd, it is unreliable.
calculating enthalpy of fusion. M, C and m,c = mass and specific heat of calorimeter and water, n, L = mass and heat of fusion of ice; T = temperature fall. 
L = (mc+MC)T/n. 
c=4.18 J/gK. assuming copper calorimeter , so C=0.385 J/gK. 
1. M = 409g, m = 45g. T = 22c, n = 14g 
L = (45*4.18+409*0.385)*22/14 = 543.0 J/g. 
2. M = 409g, m = 49g, T = 20c, n = 13g 
L = (49*4.18+409*0.385)*20/13 = 557.4 J/g. 
3. M = 409g, m = 54g, T = 20c, n = 14g 
L = (54*4.18+409*0.385)*20/14 = 547.4 J/g. 
(i) Estimate error in L from spread of 3 results. 
Average L = 549.3 J/g. 
squared differences average (variance) = (6.236^2+8.095^2+1.859^2)/3 = 35.96 
standard deviation = 5.9964 
standard error = SD/(N-1) = 5.9964/2 = 3 J/g approx. 
% error = 3/547 x 100% = 0.5%. 
(ii) Estimate error in L from accuracy of measurements: 
error in masses = +/-0.5g 
error in T = +/-0.5c 
For Trial 3 
M = 409g, error = 0.5g 
m = 463-409, error = sqrt(0.5^2+0.5^2) = 0.5*sqrt(2) 
n =(516-463)-(448-409)=14, error = 0.5*sqrt(4) = 1.0g 
K = (mc+MC)=383, error = sqrt[2*(0.5*4.18)^2+(0.5*0.385)^2] = 2.962 
L = K*T/n 
% errors are 
K: 3/383 x 100% = 0.77 
T: 0.5/20 x 100% = 2.5 
n: 1.0/14 x 100% = 7.14 
% errors in K and T are << error in n, so ignore them. 
% error in L = same as in n = 7% x 547.4 = 40
The result is (i) L= 549 +/- 3 J/g or (ii) L = 550 +/- 40 J/g. 
Both are very far above  334 J/g, so there is at least one systematic error  
e.g: calorimeter may not be copper, so C is not 0.385 J/gK. (If it was polystyrene, which absorbs/ transmits little heat, the effective value of C would be very low, reducing L.) 
Using +/- 40 is best. 
However, the spread in the actual results is much smaller
* measurements were "fiddled" to get better results; other Trials were made but only best 3 were chosen. 
<h3>Other sources of error: </h3>
L=(mc+MC)T/n is too high, so n (ice melted) may be too small, or T (temp fall) too high - why? 
* we have assumed initial and final temperature of ice was 0c, it may actually have been colder, so less ice would melt -which explain small values of n 
* some water might have been left in container when unmelted ice was weighed (eg clinging to ice) - again this could explain small n; 
* poor insulation - heat gained from surroundings, melting more ice, increasing n - but this would reduce measured L below 334 J/g not increase it. 
* calorimeter still cold from last trial when next one started, not given time to reach same temperature as water - this would reduce n.