Predicting Winners and Losers in Puppies

Much effort has been expended in attempts at identifying the puppy characteristics which predict later success as guide dogs. One approach has been to develop puppy tests which can predict training outcomes (e.g., Batt, Batt, Baguley, & McGreevy, 2008). Another approach has relied on feedback from puppy walkers (e.g., Serpell & Hsu, 2001). Each has had mixed success in predicting successful completion of training or reason for rejection (Batt, Batt, Baguley, & McGreevy, 2010). However, as yet it has not been possible to develop a tool or test that will allow guide dog schools to significantly enhance selection of puppies for dog training. Batt et al. (2010) reported that 17 of their 31 survey respondents used puppy tests. However, there were a total of 76 schools contacted in their survey, suggesting that puppy tests are not used by the majority of schools. Of the 12 participating schools in Jud Landau’s (2009) survey on puppy rejection only two indicated that they currently have formal testing in place. Each of these schools use the “Campbell test” (The Campbell test, 2010). However, Beaudet, Chalifoux, and Dallaire (1994) failed to find any predictive value for the Campbell test when used without supplementary criteria.

The absence of objective, evidence-based criteria for predicting success has meant that most guide dog schools are forced to rely on the judgments of experienced trainers when it comes to culling of pups for training. However, a minority of schools opt not to discard puppies as they believe there is insufficient evidence to do so (Jud Landau, 2009).

Given that most schools are rejecting puppies from their pool of potential guide dogs and these decisions are mainly based on the expertise of their staff, how effective are these practices? Curiously enough, there appears to be no evidence gathered to evaluate whether or not current practices add or detract value. Are trainers/puppy walking staff able to identify pups that are less likely to succeed in guide dog training?

This study examines whether or not experienced judges can predict the success or failure of puppies to complete guide dog training. Judgments were made on 50 puppies each at ages three, four, six, and nine months. Judges gave either a “Yes” or “No” on whether or not each pup would successfully complete training as a guide dog and these predictions were compared to the actual outcomes of training.

Three staff, all drawn from the Puppy Walking program, volunteered to identify pups that they believed would succeed or fail in guide dog training. Between them, the judges had accrued more than 30 years of experience in the guide dog industry. Their judgments were either a “Yes” or “No”. A total of 50 pups were judged at three, four, six, and nine months of age. The 50 pups were the first 50 pups entering puppy raising after April 2008.

Staff was not privy to each other’s judgments, nor did they keep records of their own earlier judgments. Instead, data were provided to the senior author who maintained a spreadsheet of all judgments.

Judgments were compared to the actual training outcome, that is, whether or not the pup successfully became a guide dog. This meant that the course of the study extended over approximately two years.

A statistical analysis was used to determine whether or not the judges’ predictions were any better than chance outcomes.

None of the staff predictions were better than assigning predictions (Yes/No) on a random basis as presented in Tables 1, 2 and 3.

Table 1.

Table 2.

Table 3.

When asked to predict whether or not a pup will successfully complete guide dog training, experienced staff was unable to predict any better than a coin flip. This is a sobering result in so far as the dominant practice of guide dog schools is (1) to reject some puppies from puppy walking and (2) to base the decision on the advice of staff.

Batt et al. (2010) report that use of staff feedback is associated with lower numbers of dogs completing dog training but with higher success rates among dogs placed with users. Batt et al. (2010) interprets this finding as staff successfully culling the potential poor performers so that only the better dogs are placed with users.

The Batt et al. (2010) study did not examine the accuracy of staff predictions. It simply noted whether or not staff opinion was included in the management practices of the school. Whatever the reason for fewer dogs completing training, it could be argued that this in itself would make success as a working dog more likely.

The results of the current study suggest that the practice of discarding pups on behavioural grounds is inadequate and, at the very least, should be validated by each school perhaps through the collection of hard data similar to those in the present study. As indicated in the introduction, a minority of schools do not reject at the puppy raising stage, either tacitly or explicitly recognising the lack of hard data on which to base these decisions. The results of this study are probably a bitter-sweet outcome for many schools, including Guide Dogs NSW/ACT. On the one hand, the results lend support to practices where few puppies are rejected and most assessments take place at entry to training. On the other hand, the results discourage giving weight to early judgments on puppy performance outcomes.

The current study was conducted partly in anticipation that we were able to add value by judgments at the puppy raising stage. Indeed, the next planned stage of our investigations was to discover the basis on which successful judgments were made. We have now changed our tack and are initiating research into the factors which predict success. This process will begin with a theoretical critique and lead to empirical studies over the next three years.